Shame vs Embarrassment: The Distinction In The Literature

I examine Zahavi’s paper on Shame [D Zahavi, Shame and the Exposed Self, in J Webber (ed.), Reading Sartre: On Phenomenology and Existentialism, London: Routledge, 2010] which deals with shame as discussed in the literature. Zahavi comments on analysis by Lewis, Scheler, Taylor, Harre and Mead. [Citations to follow at the appropriate point in the text.]

A principal difficulty throughout is that of distinguishing shame from embarrassment. I will show that none of the commentators discussed by Zahavi or Zahavi himself have adequate proposals to do this.


Scheler [M Scheler, Schriften aus dem Nachlass, Band I: Zur Ethik und Erkenntnislehre, Muenchen: Francke Verlag, 1957. (Originally published 1913.)] claims that blushing virgins have `protective shame’. This presumably means that on the assumption that chastity is a virtue, the shame the virgins experience when the topic of copulation is in some way brought to their attention will be of assistance in maintaining that chastity. This function would be fulfilled as well by `protective embarrassment’ and so this claim does not provide a distinguishing test.

Zahavi reports the claim made by Scheler that shame is a precondition for erotic interest in others. This could valorize it since it would be necessary for the survival of the species. Such a valorization could provide a distinction between shame and embarrassment because embarrassment either does not have value or would have a different mechanism for valorization. The mechanism proposed by Scheler for shame is that the shame associated with auto-eroticism leads the sexual impulses to turn outwards.

But we would need to have assigned a value to such survival for the link to convey any value to shame. To some extent, erotic impulses form part of the factors leading population to increase — after all, evolution has `designed’ those impulses to precisely that end. To the further extent that problems like global warming are exacerbated by population growth, then we would have further reason to assign a negative value to shame if the link claimed by Scheler holds. Any attempt to add positive value also brings with it negative value. This complexity combined with the absence of a valorization scheme for embarrassment makes it unlikely that we can find a clean distinction between shame and embarrassment from this source.


Taylor argues for a duration-based distinction between shame and embarrassment. She holds that embarrassment persists only as long as the embarrassing situation; while shame is held to persist beyond shameful events. But there seems to be no obvious reason why a memory of an episode provoking an emotion of embarrassment or shame would not provoke embarrassment or shame again on being recalled. Indeed, the ability to recall the emotion which was a fundamental aspect of the phenomenology of the event, is constitutive of having adequately recalled the event at all. The recalled emotion could potentially be in an attenuated form, but that would suffice to defeat Taylor’s claim.

Also, Taylor will have difficulties explaining why I can intelligibly make statements such as `I am embarrassed that I used to like such terrible music in college’, because there is a significant lapse of time between the state of affairs causing the embarrassment and the embarrassment itself. Her response will presumably involve the claim that the `embarrassing situation’ in this case is not the long-past enjoyment of now unfashionable music, but my current realization that in fact that music was not of the highest quality. But why is that embarrassing? Does it not on the contrary represent an admirable and not embarrassing improvement in my taste?

In addition, psychological research by Modigliani [A Modigliani, Embarrassment and Embarrassability, Sociometry, Vol. 31, No. 3 (Sep., 1968), pp. 313-326] shows that some people are more prone to embarrassment than others, and that this proneness is related to feelings of inadequacy. We know that feelings of inadequacy, or low self-esteem problems, are relatively stable since otherwise, people would not seek counseling to address these problems. This means that embarrassability must also be a stable parameter. And if that is so, then Taylor will need to explain how it is possible for some people to be more embarrassable than others when embarrassment does not persist beyond the embarrassing episode. Can it be because highly embarrassable people experience more embarrassing episodes? Or that they experience a higher intensity of embarrassment for some base-line level of embarrassment stimulus?

Zahavi gives five examples purportedly of shame: being exposed as a plagiarist, one’s friend observing the racism of one’s relative, wearing unfashionable clothes, being rejected for a job one expected to receive, not wearing make-up. This illustrates the prevailing predicament in that all of these could equally well be described as embarrassment and we would have no difficulty accepting a spread between the two emotions, if indeed two they be.

One cannot be ashamed of events for which one is not responsible. The racism example looks like it could be a counter-example to this. It is clear that no-one can take a great deal of responsibility for the political views of relations. But it might be the case that the shame derives from a failure to prevent the event from taking place. Left unspoken in Zahavi’s example is the possibility that the friend with one is in fact a member of an ethnic minority, and was the target of the racism. There, one would certainly feel responsible for having allowed the situation to take place.

Zahavi also claims that it is implausible that any of these scenarios could have happened when alone — but does that commit him also to saying there cannot be shameful memories? He may have a counter to the extent that one could feel shame on recalling a shameful memory not in virtue of that recall but in virtue of the recalled possibility of further exposure of the remembered episode in the future to other observers. We then have to decide to what extent we are `alone’ with our memories — would someone irrevocably marooned alone on a desert island have shameful memories? Zahavi has to say no, but if we imagine that our castaway has in fact been exiled for heinous crimes of which he now repents, Zahavi’s position appears dramatically implausible. So Zahavi’s various claims on shame are questionable and do not provide a distinction between shame and embarrassment.


Zahavi discusses Harre’s claim [R Harr\’e, Embarrassment: A conceptual Analysis, in W Crozier (ed.) Shyness and Embarrassment: Perspectives from social psychology, Cambridge: Cambridge University Press, 1990] that the distinction between shame and embarrassment is that in the former case, others see our moral breach while in the latter case, others see our breach of convention. Zahavi criticizes this for being too neat: it has too clean a division between moral and conventional infractions. This crticism has force, not least because the purported distinction assumes that morality and convention are separate. Against that, many philosophers argue that in fact morality is merely conventional. A plethora of citations could be given to illustrate this: I will restrict myself to Sedgwick’s [P Sedgwick, Nietzsche: the key concepts, Abingdon: Routledge, 2009, p. 100] observation on Nietzsche: `[t]he claim that morality was, in the first place, simply obedience to customs and no more is fruitful’.

Zahavi proposes an alternative distinction: shame is unlike embarrassment in that only the former is accompanied by a decrease in self-esteem. But embarrassment is undoubtedly not a positive emotion. If we contemplate a case of extreme, overwhelming, all-encompassing embarrassment, is it likely that it would have no negative effect on our view of ourselves? Would we not think the mere liability to feel such an emotion to such a paralyzing extent reflected badly on ourselves, not least because we might be prevented from acting in productive ways by our feelings? In any case, Zahavi is here making a false empirical claim. Modigliani relates susceptibility to embarrassment to inadequacy, describing how `the individual may suffer the relatively severe loss of situational-self-esteem associated with embarrassment by failing to manifest situationally appropriate demeanor’. Situational-self-esteem is that element of self-esteem that derives from behaving appropriately; we are all aware of the perverse tendency to giggle at funerals.

However, Modigliani allows that the effect may be transient; the embarrassed individual `is likely to find that his general self-esteem has been little affected by the experience’, [ibid., p. 315] but Zahavi still has questions to answer. These questions are made only more pointed by the examples of embarrassing situations given by Modigliani: `e.g. a slip of the tongue, a revelation of negative self-information by others, an inability to carry off situationally relevant tasks’ [ibid., p. 316], as all of these are clearly embarrassment and not shame and all of them will indubitably cause a decrease in self-esteem. How could they not? They are failings. The objection that Modigliani has himself confused shame and embarrassment is scarcely available to those arguing that the self-esteem decrement is definitional of shame, unless they also claim that it is not a definition that is widely known.

But worse for Zahavi is that there are even physiological stress responses associated with both shame and embarrassment, and that these are similar for both emotions. Lewis and Ramsay [M Lewis and D Ramsay, Cortisol Response to Embarrassment and Shame, Child Development Vol. 73, No. 4 (Jul. – Aug., 2002), pp. 1034-1045] measure cortisol levels in children that have been placed in embarrassing and shameful situations. Cortisol is a hormone released in response to stress. [G Baker et al, \underline{Stress, Cortisol Concentrations, And Lymphocyte Subpopulations}, British Medical Journal, Vol. 290, No. 6479 (May 11, 1985), p. 1393] Lewis and Ramsay found that `cortisol response scores for the children who showed [….] embarrassment and for the children who showed shame’ were very similar `indicating that the relation to cortisol response was comparable for the two emotions’. If the stress is not related to a decrease in self-esteem, then where is it from?

In addition, Zahavi may be challenged from a gender perspective. Lewis and Ramsay note that `boys show more exposure embarrassment than girls did at 22 months’ while `girls consistently show more shame than do boys’. To be fair, the authors also note that the embarrassment difference fades by 35 months. Nevertheless, Zahavi is committed by this evidence to the claim that girls have lower self-esteem than boys. That claim may have some currency, but as Gentile [K Gentile, Review: [untitled], NWSA Journal, Vol. 3, No. 3 (Autumn, 1991), pp. 481-483] notes in a review of work by Kenway and Willis [J Kenway and S Willis, Hearts and Minds: Self-Esteem and the Schooling of Girls, New York: Falmer Press, 1990], `self-esteem studies […] surprisingly, find no empirical evidence to support the wide-spread notion that girls are deficient in self-esteem’.

The other side of the distinction that Zahavi needs is that shame always causes a decrease in self-esteem. But imagine some group whose members gain approbation within the group from acts that non-members of the group would call shameful. For example, for a prospective recruit, `murder would form part of a gang initiation rite’. [BBC Reporter, Race killer’s prison boast, London: British Broadcasting Corporation, 19 February 1999] No doubt the recruit would not feel shame and could indeed feel an increase in self-esteem at becoming a gang member. That is fine for Zahavi, but he needs to explain the sequence of events if subsequently the gang member repents of his actions, which would involve shame caused by a change in the accepted relevant group of evaluators, being wider society rather than the gang. Zahavi needs this penitent to feel a decrease in self-esteem, but would he not on the contrary feel an increase because he had now realized his errors and become rehabilitated in wider society? If Zahavi counters that the self-esteem increment derives from the rehabilitation and the shame from the belated realization that the original action was in fact shameful, he will need an analysis that separates two elements that at least in this story are inseparable: the gang member cannot have the rehabilitation if there is nothing from which to be rehabilitated.

In support of his line, Zahavi notes a remark he ascribes to G Strawson that past episodes of embarrassment can be retailed because they are funny but not past episodes of shame: these continue to be shameful and will remain hidden. This can be questioned. Although there are many examples of people telling stories about situations in which they were embarrassed at the time, is it really the case that no-one ever tells such a story about a shameful situation, no matter how long afterwards? Zahavi may counter that it must be an amusing story, but even that seems possible. There is a story in my family relating to a wartime exploit involving my grandfather, a large mobile anti-aircraft gun and the destruction of civilian property. This would doubtless have been an occasion for shame rather than embarrassment at the time, but the story was nevertheless extant during the life of my grandfather. Zahavi has a further counter about time expired, but his defense would then be ad hoc. Since the exercise aimed to construct a clean distinction, it will have failed if it relies on vague concepts like `sufficient’ elapsed time.


Zahavi discusses Mead [G Mead, Mind, Self and Society, Chicago: University of Chicago Press, 1962] in the context of attempting to elucidate what shame tells us about the self. Zahavi’s own view is that `[t]he shamed self is a more complex (and complicated) self than the experiential self’, meaning that shame is necessarily social and experience is not. If this is intended as a potential distinction between shame and embarrassment, then again, Zahavi has empirical problems. As Tangney et al [J Tangney and D Mashek, Shame, Guilt, and Embarrassment: Will the Real Emotion Please Stand Up?, Psychological Inquiry, Vol. 16, No. 1 (2005), pp. 44-48] show, embarrassment is more socially oriented than shame: `when embarrassed, people were more likely to feel that others were looking’; moreover, `the shamed individual feels isolated from others’.

Zahavi describes Mead’s view that self-consciousness is a by-product of becoming aware of others; the idea being that others are initially more salient to the infant than the infant is to itself. They would then become aware of themselves — `passing the mirror test’ — and note the physical similarity between themselves and others as an intermediate step to postulating that they are mentally similar to others, once they also observe that others seem to have mental states and desires.

The standard — but not universally held — view in psychological research is that infants develop a theory of mind between the ages of three and five. [K Riggs (ed.), Children’s Reasoning And The Mind}, Hove: Psychology Press Ltd., 2000] Prior to this point, they fail the false belief test: they assume that if they know something then everyone does. Given that, if Mead is correct that self-consciousness requires awareness of others, then there could be no self-consciousness in infants prior to the ages of three to five. This would support my previous contention that the evidence for phenomenal awareness in infants is slimmer than Zahavi supposes. Mead holds that prior to the advent of self-consciousness, feelings would be experienced as part of the environment. That also supports my view that infants do not have phenomenal experience because otherwise we would have to say that `there is something it is like for them’ and that that something is `external to them’.

Davidson on the Structure And Content Of Truth

Davidson: `Structure And Content Of Truth’

Davidson cites Tarski (p. 291); Tarski holds that any semantic conception of truth must diverge from its meaning on a `meaning is use’ conception, since `the common meaning of the word “true” — as that of any other word of everyday language — is to some extent vague.’ This may be questioned in a variety of ways. Firstly, we may ask whether “true” really is vague in everyday use. Certainly that will be the case if by the claim we mean that ordinary speakers have not decided whether they prefer a correspondence, pragmatist or other theory of truth as they clearly will not have done. But presumably Tarksi means something more: that there will be indeterminate borderline cases where it is unclear to ordinary speakers whether what they say is true or not. This may in fact be the case, but it will not appear so to speakers. They will know when they are unsure of something. But that will just represent, for example, epistemic uncertainty — not any belief on the part of the users of the term true that it is somehow fuzzy at the edges. Secondly, we may ask whether an examination of ordinary usage is really the best way to examine the notion of truth or whether the attempt is better conducted — at least initially — on a more restricted and well-behaved formal language. Since Tarski may be roughly characterized as favoring a `redundancy’ model of truth — viz. that the assertion `It is true that A’ does not say more than `A’ — he may not believe that attempt is worthwhile. Finally, Tarski assumes that if the vagueness he describes is really present, ineliminable, and a part of the most useful language to be analyzed, that such vagueness cannot be accommodated in the theory. And as Davidson points out (p. 294), Tarski himself claimed to have ` “caught the actual meaning” of the intuitive concept of truth’ — so either his theory accommodates this vagueness or it was not present to be accommodated.

Later (p. 298), Davidson criticizes both epistemic and realist views of truth — the former holding that truth is in some way mediated by finite human capacities, and the latter denying this. Davidson’s criticism is that both approaches `invite skepticism’. This should not be used as a criticism of theories, in the same way that Nozick was mistaken in presenting the main benefit of his tracking theory of truth as a defense against skepticism. His theory had value without that. Since few of us take the skeptical challenge seriously, avoiding or defeating it should count so much the less in assessing the merits of competing theories. As Nozick said, the real interest in skepticism is his formulation `how is knowledge possible?’ rather than `could we all really be brains in vats?’

What is an empirical theory of truth?

  • Starting point: looking for a theory of meaning
  • Motivation: cannot base theory of meaning on the form `s means m’ because `means that’ is an intensional context → logically difficult to analyze
  • Also difficult to find singular terms in `m’ to refer to meaning


Davidson’s approach to theory of meaning

  • Equivalent sentences: examine `s means p’ → `p’ is another sentence
  • Seek `matching sentence’ to replace `p’ which `gives meaning’ of `s’
  • `Bold step’: make the `p’ position extensional; make three changes
  • Eliminate `means that’ because non-extensional
  • Prefix `p’ with sentential connective so we can analyze the logic
  • Apply a predicate to `s’
  • Result → (T): s is T if and only if p
  • Also: require T predicate to entail all sentences when `s’ becomes description of sentence and `p’ becomes same sentence
  • But this is just Tarski’s convention T!


Convention T

  • T introduced by Tarski in `disquotational’ theory of truth
  • Example (E): “snow is white” is true if and only if snow is white
  • Use/mention distinction prevents circularity from multiple occurrences of e.g. `snow’
  • Problem: also works with `grass is green’ on RHS since we are only interested in truth conditions
  • Response: if this were part of a theory of truth also including `that is snow’; truth conditions would no longer match
  • Also, T sentences are supposed to be law-like



  • Theory of truth can be `foundation’ of theory of meaning
  • Theory of meaning empirical; account for natural languages
  • Test theory by comparing predictions with facts
  • Here, that means seeing whether E holds


Empirical tests

  • How do we verify a theory of truth empirically?
  • Check that all the T sentences, or a sufficiently large sample, are true
  • We only have access to behavior including utterances of speakers
  • Adapt Quine’s notions of radical translation and matching to behavior/speech
  • Find out what sentences speakers accept as true
  • GE is evidence for T:
  • (T) `Es regnet’ is true-in-German when spoken by x at time t if and only if it is raining near x at t
  • (GE) (∀x)(∀t) (if x belongs to the German speech community then (x holds true `Es regnet’ at t if and only if it is raining near x at t’))
  • Problems with extensive nature of (∀x)(∀t)
  • Problems with error: speakers may be wrong about whether it is raining or not
  • (GE) is only supposed to be a generalization: not always true
  • Seek `best fit’; `maximize agreement’ making speakers right as much as possible
  • Something like a Principle Of Charity — but based on the claim that if we cannot find agreement, then speakers are not rational



  • Tarski’s analysis starts from the observation that the liar paradox is a feature of all sufficiently rich languages
  • Addresses this by working in a metalanguage on a more restricted formal language
  • This can be done in stages but will mean no adequate truth theory for natural languages is possible; the restricted target languages will just look like formal languages
  • Gödel’s result on the non-coextensivity of provable and true statements in arithmetic derives from the liar paradox ultimately
  • Arithmetic a simple formal language — shows Tarski’s concern genuine if Gödel’s result means there can be no theory of truth in arithmetic
  • Tarski’s view described as a `redundancy’ theory of truth: there is no difference between asserting `It is true that A’ and simply `A’

Quine And Fine on Reference and Modality

Quine: `Reference And Modality’

Quine argues that there is no inference from `There is no such thing as Pegasus’ to `(∃ x) (There is no such thing as x)’, which is true. And yet, there is such a generalization from `There is no reference for the term Pegasus’ to `(∃ x) (There is no reference for the term x)’. Since `Pegasus’ does not refer, use of the term can only refer to the name itself. The name can have properties even though the referent of the name, being non-existent, can not.

Similarly, one can question another claim of Quine’s that one exception to existential generalization is given by noting that it is not possible to generalize from `Giorgione was so-called because of his size’ to `(∃ x) (x was so-called because of its size)’. This seems to rely too much on the precise usage of `so-called’, which does indeed, as Quine points out, lack a defining antecedent in the generalization given. Yet there is a perfectly good generalization to `(∃ x) (x was called x because of its size)’, which captures the meaning well, suggesting that Quine has been over-reliant on the properties of `so-called’ here.

If these two substitutions are innocuous in a Wittgensteinian sense such that meaning is use, then their inter-substitution should not change the results that Quine seeks.

Why is `the number of the planets’ a name of the number 9? Surely Quine’s point is that the number nine has certain properties necessarily, while the number of planets is contingent, as indeed demonstrated by the demotion of Pluto from planetary status. Presumably Quine will respond that there is no need for a name to be linked necessarily to its referent. But that merely plays on the unhelpful trope that `9′ could refer to a different number — such that `9′ could be less than 8, if `9′ referred to the number we currently refer to with `2′.

Fine: ‘The Problem Of De Re Modality’

Fine disagrees (p. 218) with Quine’s claim that failure of substitutivity of two terms suffices to show that they are not co-referential. One of his arguments for this is that it is not possible to substitute a co-referring substitute for `nine’ into `canine’ within the laws of grammar. This is true, but the failure of substitutivity is of a different type to the one Quine considers. Quite simply, `can[X]’ is not a word in English for almost all values of X. It is this which prevents the substitution — it is generally not permitted to break words up and make arbitrary replacements of some letters while retaining the original reference or indeed any reference at all. So Fine is wrong to say that this type of failure of substitutivity is a counter-example to Quine’s rule that failure of substitutivity entails failure of co-reference.

Fine’s second complaint here is that a language may be `impoverished’ such that there is no term co-referential with `nine’ that we may use to check for failure of substitutivity and to see whether failure of co-referentiality is invariably entailed. Yet this absence is purely contingent. If we define the neologism `morgon’ to be co-referential with `nine’, we may ask the question whether substitution changes truth value. This seems to be the case because “ `Nine’ has four letters” is true while “ `Morgon’ has four letters’ is false. Yet Quine will surely respond that these quotational contexts are referentially opaque and thus not open for substitution.

Fine addresses this response by allowing his opponents to require that the new sentence after substitution is in fact a sentence according to the language. He allows that this solves the first difficulty but not the second. His preferred solution is to restrict his analysis to languages which do not prevent such substitutions. But then we are no longer talking about natural languages.

Fine’s further complaint is that substitutivity cannot be examined in sentences which are disjuncts with a logical truth. For example, we cannot substitute `Giorgione’ in the sentence “ `Giorgione’ was so called because of his size or 2 + 2 = 4′ to check for changes in truth value because the disjunction is always true. In fact, this is the case whatever we substitute for `Giorgione’. And yet, is this not because in fact we do not consider — or need to consider, at least — the first clause at all? What would we say if the first clause became meaningless, or contained non-referring terms? We would probably conclude that the sentence as a whole remains true, thus showing that in fact the structure has merely removed the first clause from considerations altogether. That is why it fails as a substitutivity test: the relevant element is no longer under consideration.