What Emotions Really Are: The Problem of Psychological Categories

paul e. griffiths

Science and its Conceptual Foundations

Chicago and London: The University of Chicago Press 1997, xi+286 pp, US$27.50 Hb.

This pithy book is for any psychologist or philosopher who wants to do psychology in a biologically informed way. Emotions are an object lesson, and the lesson is mostly negative: emotions are no one thing, and most of them are something we know not what.

Griffiths is a sure footed guide through some of the most rocky terrain in contemporary philosophy of biology: the subtle pitfalls of what's left of the nature nurture question; the best conception of natural kinds outside of physics, and the troubles of the adaptationist program on which recent controversies have thrown more heat than light. Griffiths also shows how to fix these last problems, by paying careful attention to matters of lineage now accessible through DNA analysis (though he writes, rather oddly, as if only strict cladists took lineage seriously). While studies of ancestral relatedness may not bring definitive confirmation of adaptationist hypotheses, they can invalidate them if they turn out to require that A come about only in the context of B, where B turns out to be nonexistent at the time of A's appearance. There is also an excellent discussion of the various levels at which psychological and biological phenomena might be studied. Griffiths's version of this hierarchy of explanation distinguishes an ecological level, which describes the function or goal of the trait at the most general level, a computational level, which specifies the strategy for the attainment of the goal, and an implementational level, at which the causal nitty gritty actually takes place. Any projectable concept can ground a useful natural kind, but only the causal nitty-gritty counts as an acceptable basis for forming a projectable concept, and only genuine homologies grounded in cladistic analysis can establish that we have a proper grasp of that level. This makes Griffiths's requirements for an acceptable theory of what "something really is" very demanding. He is not content with any functional or ecological explanation unless the full causal story of its implementation can also be told. What this means for emotions is that they do not exist -- or rather, there are at least two and perhaps three sorts of things we call "emotions," but only one -- the quasi-automatic response routines encompassed by the "affect programs" of Paul Ekman and others -- meets those stringent requirements. Of the others, more in a moment.

Griffiths is sneeringly contemptuous of "the philosophers'" method of conceptual analysis, which he associates with equally useless variants of what he calls "propositional attitude" theories. He accuses the philosophers of ignoring the progress made in the last few decades in the philosophy of language and philosophy of science. This harsh judgment is not altogether undeserved: many philosophers seem not to have noticed the demise of the analytic-synthetic distinction, or the possibility of natural kinds that can do useful jobs without being ultimate repositories of natural law. The method of conceptual analysis might do for a lexicographer, but is entirely the wrong method for arriving at a substantive definition of what something really is. If it were not so, then an adequate theory of phlogiston could have been found by just listening carefully to what people said about it. As it is, the uses of the word 'emotion' are driven in part by "nonepistemic dynamics," with the result that "[u]sing these concepts as a guide to the emotions is like studying female sexuality by reading pornography." (201)

But what if what emotions "really are", in this rigorous biological sense, is not really what they are? What if they are really no more biologically real than legal systems, say? In reply, Griffiths has a very good chapter on "social constructionism". He disentangles the various meanings of this doctrine. One is the "social concept" theory which classifies emotions as incommensurably variable across cultures because their objects are "irreducibly socio-cultural." This view he dismisses, mainly on the ground that it is committed to an absurdly strong holism. He draws blood when he points out that philosophers who have "spent their careers denouncing" the view that emotions are passive as nothing but a cultural myth are ill-equipped to proclaim that emotions just are, after all, nothing more than cultural myths. (149) He argues that the only interesting construal of social constructionism is that variant of the "social role" theory according to which emotions are "covert constructions." This makes them, in effect, "disclaimed actions" by agents who take them to be passions, and who are therefore systematically self-deceived. Griffiths plausibly insists that not all emotions could be disclaimed actions. This is surely right, but one might object that he grants the social constructionist too much: even according to the vernacular no true emotions are disclaimed actions. Those disclaimed actions -- running amok, behaving angrily to secure some unconscious goal -- express only phony emotions, drawing their credentials from the way they mimic real ones drawn from the repertoire of Griffiths's "affect program emotions."

Many philosophers have repeated, like a mantra, the view that emotions do not form a natural kind. But -- and here I must plead guilty myself -- they have gone on to lump them together anyway. Griffiths actually takes this denial seriously, and makes it interesting by sharply distinguishing three distinct classes among what are called 'emotions'. The first class comprises emotions which do form a natural kind, or maybe six natural kinds: these are the ones that are implemented as the "affect programs." They are relatively modular, and they have probably identifiable neural bases and psycho-evolutionary origins. They are surprise, fear, anger, disgust, sadness, and joy. Another class of emotions, Griffiths calls them "higher cognitive emotions," are more responsive to cultural and cognitive influence; to a third kind, the social constructionist view actually applies: as disclaimed actions they are in fact "essentially pretenses," and so "they can be sincere only in the way that... ghost possession can be sincere." (158)

Most of this is invigoratingly persuasive. But the hapless philosophers thus pinned to the ground might manage a few squeaks of protest.

Griffiths's contrast between the affect program emotions and the "cognitively penetrable" class may be too sharp, for two reasons. First, if one places the cognitive class of emotions at too great a distance from the physiological mechanisms of the affect program emotions, the old problem of "unemotional evaluations," which Griffiths rightly lists as one of those plaguing the "propositional attitude approach"(29) will recur. This may not bother Griffiths, however, since he himself does not claim to have any theory of those "higher cognitive emotions." Second, while emotions of this last class are not strictly modular, the notorious single-mindedness of even cold anger, jealousy, or fear concentrates the mind in a way that seems to mimic the modularity of sensory modalities. This resemblance may suggest that there is more descriptive and explanatory juice to be squeezed out of the functional unity of emotions, both hot and cold, than Griffiths's stern requirement of cladistically validated homology allows. But Griffiths does not altogether deny this either: he just views a unity of ecological function as relevant to only half the task of explanation. Only when we know how this function is implemented will we be entitled to boast that we have the other half to make a real explanation.

Lastly, Griffiths's attack on the "propositional attitude" or "cognitive view," like the defense of that view by most of the philosophers who proclaim allegiance to it, suffers from the inherent vagueness of the notion of cognition. The classic refutation of the "cognitivist" view is in a famous 1980 paper by Zajonc (Zajonc 1980) which reports some ingenious experiments in which priming by exposure to tunes or images produces preference without any ability to recall them, thus showing, in Griffiths's words, "an apparent emotional response but no evidence of higher cognitive activity." (93). This "proof" seems to confuse cognition with awareness, insofar as awareness is linked to recall ability. But by that criterion one could also show that there can be finely turned motor activity without awareness. For according to some findings by Melvin Goodale, "[e]vidence from both neurological patients and normal observers suggests that many of our actions are controlled by visual inputs that are quite inaccessible to conscious experience" (Goodale 1998). It seems reasonable to characterize cognition broadly as the assimilation of information relevant to thought, feeling or action; in which case there is no proof here of the irrelevance of cognition. Indeed, as Zajonc's title suggested, it is inference, not cognition, which he purported to show was not needed for preference. (In addition, an emotional cognitivist might well claim that preferences, like startle (cf. 241), are too simple to be emotions even on the affect-program level.) But although this problem about the interpretation of the "cognitive" is real, I don't think it affects anything important in Griffiths's argument.

His take-home message seems to me provocative and profound, as well as probably correct. Emotion (and probably much else in our mental life) is a "heterogeneous construction" (132) from genetic and environmental inputs which interact in a "non-additive" way. The results usually look smoothly adaptive, but their causal underpinnings are piecemeal, to an extent that the apparent coherence of our responses makes it hard to realize. They have been assembled from heterogeneous bits and pieces with the opportunism characteristic of evolution. This tells us much about psychology, and about the ultimate futility of our attempts to construct a rational model of a rational agent. Read the book.

ronald de sousa University of Toronto


Zajonc, R. B. 1980. "Feeling And Thinking: Preferences Need No Inferences." American Psychologist 35:151-75.

Goodale, Melvyn. 1998. "The Zombie Within: Evidence for Vision Without Sight in the Normal Observer." Paper presented at a conference on Naturalism, Evolution, and Intentionality at the University of Western Ontario, March 1998.