Graham Davey
It is over 20 years since Seligman (1970, 1971) introduced the concept of biological preparedness to explain why fears and phobias are so much more likely with certain stimuli (e.g. snakes, spiders) than others, and a glance at contemporary psychology text books suggests that his theory has stood the test of time. The past ten years, however, have seen a significant increase in our understanding of cognitive biases in the processing of threatening stimuli. This target article will argue that such biases can also explain most of the phenomena traditionally attributed to phylogenetically based predispositions to associate fear-relevant stimuli with aversive outcomes. The paper consists of (i) a description of the phenomena to be explained, (ii) an evaluation of the evidence favouring either biological preparedness or cognitive biases, and (iii) a description of the evaluative processes that might contribute to an information processing explanation.
1. THE PHENOMENON TO BE EXPLAINED
Explanations in terms of conditioning treat all stimuli as equally likely to enter into association with aversive consequences, yet fears and phobias are not evenly distributed across stimuli and experiences. The evidence comes from a number of surveys of fears and phobias reported in the literature (e.g. Agras, Sylvester & Oliveau, 1969; Costello, 1982; Kirkpatrick, 1984; Bennett-Levy & Marteau, 1984). Consistent across these surveys are the types of stimuli that are most likely to elicit fear. These include animals, particularly snakes and spiders, heights, thunder, fire, deep water, and death. These fears all appear to have prima facie evolutionary significance in that each would have represented survival threats to pretechnological people (e.g. Seligman, 1971). Laboratory studies suggest that fear is acquired more readily with these stimuli than with less feared stimuli (e.g. Mineka, 1987) and that it is also more resistant to extinction (e.g. McNally, 1987).
2. EVOLUTIONARY vs. MECHANISTIC EXPLANATIONS OF THE UNEVEN DISTRIBUTION OF FEARS.
Behavioural phenomena can be explained in two different and relatively independent ways: in terms of the evolutionary pressures that selected for them or the biological or psychological mechanisms that mediate them, the former is known as an 'ultimate' explanation and the latter as a 'proximate' one. These two types of explanation are relatively independent because ultimate explanations do not presuppose a particular kind of proximate explanation, largely because evolutionary pressures select for outcomes and not mechanisms (cf. Plotkin & Odling-Smee, 1982; Plotkin, 1983).
Traditional explanations of the uneven distribution of fears have focused on the role of evolutionary pressures in selecting for associative predispositions with fear-relevant stimuli. For example, Seligman (1970) proposed that "The organism may be more or less prepared by the evolution of its species to associate a given CS and US or a given response with an outcome" (1970, p408). Hence, the fact that snakes represented a critical danger for pretechnological people selected for a predisposition to learn rapidly to associate snakes with aversive consequences. Similar effects occurred with other stimuli such as spiders, fire, deep water, etc.
"Biological preparedness" hypotheses are primarily ultimate explanations, accounting for the uneven distribution of fears in terms of original selection pressures, but there has also been an attempt to specify at least some of the details of the proximate mechanisms: associative learning is facilitated when an aversive consequence is paired with a "fear-relevant stimulus".
3. BIOLOGICAL PREPAREDNESS AS AN EXPLANATION OF THE UNEVEN DISTRIBUTION OF FEARS.
3.1 What is the biological preparedness hypothesis?
According to Seligman, associative learning is not an arbitrary process; the ease with which learned associations can be formed will depend on biological predispositions shaped by the specialised evolutionary history of a species. Although this is an evolutionary view of learning, Seligman operationally defined preparedness quite independently of this view: "The relative preparedness of an organism for learning about a situation is defined by the amount of input (e.g., number of trials, pairings, bits of information, etc.) which must occur before that output (responses, acts, repertoire, etc.), which is construed as evidence of acquisition, reliably occurs." (1970, p408). Using this definition, Seligman then went on to point out that an organism can be either prepared, nonprepared or contraprepared for learning. Preparedness is a predisposition to learn an association rapidly; nonpreparedness is the neutral middle of the preparedness continuum, where the associations to be learned are acquired neither rapidly nor slowly; and contrapreparedness is the opposite pole, where an animal learns an association only with great difficulty.
3.2 Applying biological preparedness to an understanding of human fears and phobias
In 1971, Seligman extended his preparedness thesis to human fears and phobias. At the time, this seemed a logical extension of his ideas and gained further influence because other explanations of phobias were much less successful in accounting for all the relevant phenomena. Human fears appeared to be "(1) selective, (2) ... resistant to extinction, (3) irrational, and (4) capable of being learned in one trial." (1971, p312). None of these facts was compatible with traditional conditioning accounts of human phobias based on contiguity between the phobic stimulus and an aversive outcome (cf. Marks, 1976). In particular, "..a neglected fact about phobias is that, by and large, they comprise a relatively nonarbitrary and limited set of objects: agoraphobia, fear of specific animals, insect phobias, fear of heights, and fear of the dark, etc. All these are relatively common phobias. And only rarely, if ever, do we have pajama phobias, grass phobias, electric-outlet phobias, hammer phobias, even though these things are likely to be associated with trauma in our world" (1971, p312). Seligman suggested that "phobias are highly prepared to be learned by humans, and, like other highly prepared relationships, they are selective and resistant to extinction, learned even with degraded input, and probably are noncognitive." (1971, p312). He went on to link this preparedness to evolutionary origins, pointing out that the great majority of phobias are about objects of ecological importance to the survival of the species (e.g. fear of darkness, fear of dangerous animals such as snakes and spiders).
3.3. Defining preparedness
In his 1970 paper, Seligman wrote that a prepared stimulus "is defined by the amount of input..which must occur before that output..(which is construed as evidence of acquisition) reliably occurs" (1970, p409). More recently, however, the terms "fear-relevant" and "fear- irrelevant" have tended to be substituted for "prepared" and "unprepared" (e.g. !hman, Dimberg & !st, 1985) because they are theoretically neutral in that a stimulus is defined as 'fear-relevant' merely on the basis of its frequency in the general population as a fear-evoking stimulus, or on the basis of ratings of its fear-evoking properties by normal subjects (e.g. Bennett-Levy & Marteau, 1984). This neutral terminology (henceforth abbreviated FR and FI) will be used in the remainder of this article; the terms "prepared" and "nonprepared" will only be used when their original theoretical meanings are implied. The rapid learning and resistance to extinction usually found with FR stimuli have traditionally been known as "preparedness effects". To avoid theoretical connotations, these effects will be called "selective associations".
3.4 Evaluation of the biological preparedness hypothesis
At the proximate level, the biological preparedness hypothesis predicts that so-called prepared fears will exhibit the dynamic properties (i.e. rapid acquisition, resistance to extinction) characteristic of phylogenetically facilitated associations (cf. Seligman, 1971). At the ultimate level, the preparedness hypothesis can be tested in circumstances where the subject's experience of prepared stimuli has been strictly controlled. Subsequent first-time exposure to prepared stimuli in a learning task should still result in rapid acquisition of fear toward the prepared stimulus (e.g. Mineka, 1987). In addition, those stimuli that were directly involved in the selection pressures that shaped prepared associations ("phylogenetic stimuli") can be compared with stimuli which could potentially exert selection pressure but are too recent in the history and experiences of humans to have generated genetically mediated predispositions (henceforth these will be called "ontogenetic stimuli", e.g. guns, electricity outlets). Showing that phylogenetic and ontogenetic stimuli exhibit similar selective association effects, however, does not falsify the biological preparedness hypothesis; it merely makes the account less parsimonious. In evolutionary terms, imparsimony is not necessarily damaging because many different mechanisms might have evolved separately to produce similar behavioral outcomes.
3.4.1 Evidence from laboratory studies of human subjects
Much of the early enthusiasm for biological preparedness resulted from the pioneering laboratory work of Arne !hman and his colleagues (cf. !hman, 1979; !hman, Dimberg & !st, 1985, for reviews). They used slides of snakes and spiders as FR conditioned stimuli (CSs) and paired them with electric shock as the unconditioned stimulus (UCS). Conditioning to these FR stimuli was then compared with conditioning to FI stimuli such as slides of houses, flowers, and mushrooms. An exhaustive review of these early laboratory studies by McNally (1987) concluded that although there was firm evidence for enhanced resistance to extinction of fear responses conditioned to FR stimuli, evidence for rapid acquisition and the resistance to instructions of selective associations was, at best, equivocal. Much has already been written about these studies and the reader is referred to McNally (1987), Mineka (1985), !hman, Dimberg & !st (1985), and LoLordo & Droungas (1989) for discussion of them.
3.4.2 Evidence from clinical studies
Clinical studies have attempted to provide evidence for preparedness theory by demonstrating that the majority of clinical phobias are fears that can be classed as "prepared" and as such exhibit the dynamic characteristics of prepared stimuli (e.g. rapid acquisition, resistance to treatment, etc.).
Before we evaluate this clinical approach, note that investigators were initially confronted with a definitional problem. Seligman's original definition of prepared phobias was unhelpful because he defined prepared stimuli in terms of the dynamics of the acquisition and extinction process associated with them. Hence, because clinicians wanted to see whether phobias classed as prepared exhibited these dynamic characteristics, they could not use these characteristics to define them. Investigators tried to overcome the problem by asking expert raters (all of whom had a basic knowledge of Darwinian theory) to define the extent to which they thought various fear stimuli might have been dangerous to pretechnological people (e.g. deSilva, Rachman & Seligman, 1977; Zafiropoulou & McPherson, 1986: deSilva, 1988; Merckelbach, Van den Hout, Hoekstra & Van Oppen, 1988). It was assumed that these ratings would help define prepared stimuli per se, whereas in reality they could at best only identify the selection pressures that led to the evolution of prepared associations. For example, it is improbable that predatory pressures would select for fear toward specific predators because the variety of animals that would prey on a specific organism is likely to change more rapidly than the time it would take for the fear of specific predators to be encoded in the gene pool. Predatory pressures would be more likely to select for fear of relatively invariant predatory stimulus configurations such as rapid approach, looming shadows, being stared at or followed, etc. (see Russell, 1979, for a fuller discussion of these characteristics).
As an illustration of the doubtful validity of the expert rater method of defining prepared stimuli, consider fear of riding in a car: One might suppose that this could not have arisen from specific evolutionary selection and would hence be nonprepared (see deSilva, Rachman & Seligman, 1977), but it could be mediated by a fear of rapid movement towards the individual (e.g. an oncoming car) - a specific configuration that might well have been selected by evolutionary pressures.
Even apart from these criticisms of the expert rater method, clinical support for biological preparedness theory has not been very compelling. First, these studies have not always confirmed that prepared fears represent the majority of clinical fears. Studies by deSilva, Rachman & Seligman (1977), Zafiropoulou & McPherson (1986) and deSilva (1988) support this hypothesis, but Merckelbach, van den Hout, Hoekstra & van Oppen (1988) found that most of the clinical phobias in their sample were rated as nonprepared rather than prepared.
The clinical studies went on to compare the characteristics and dynamics of phobias rated as prepared and nonprepared. All failed to support predictions derived from Seligman's (1971) original preparedness hypothesis. Degree of preparedness was found to be unrelated to outcome of therapy, duration of phobia, suddenness of onset, severity of impairment, intensiveness of treatment received, age of onset, and impaired reproductive capacity. Only the study by Merckelbach et al. (1988) offered any support for preparedness predictions by demonstrating that in obsessive-compulsive patients prepared fears were associated with relatively poor treatment outcome.
In summary, the evidence from clinical studies largely fails to support predictions from biological preparedness, leading some writers to note the limitations of preparedness as an explanation of clinical phobias (e.g. de Silva, Rachman & Seligman, 1977; Zafiropoulou & McPherson, 1986). To be fair to the biological preparedness account, however, clinical phobias may not be the best examples on which to test the theory. Since clinical phobias are often the severest forms of fear, studies which attempt to demonstrate dynamic differences between prepared and nonprepared phobias may be limited by ceiling effects. In addition, problems in defining prepared phobias within the clinical paradigm may have added to the difficulties inherent in such studies, making it difficult to confirm the preparedness hypothesis.
3.4.3 Evidence from primate studies
Some of the most compelling evidence for biological preparedness has come from studies of the observational learning of fear of snakes in primates (rhesus monkeys, Macaca Mulatta) by Susan Mineka and her colleagues (Mineka, 1987). Naive laboratory-bred observer monkeys view a video of an experienced demonstrator behaving fearfully in the presence of a variety of FR or FI stimuli. The observer monkey is subsequently exposed to these stimuli and his reaction to them is analysed for evidence of acquired fear. These studies have indicated that monkeys who are not initially afraid of snakes will rapidly acquire an intense fear when they have watched a wild-reared monkey behaving fearfully in response to a toy snake (Cook, Mineka, Wolkenstein & Laitsch, 1985; Mineka, Davidson, Cook & Weir, 1984; Cook & Mineka, 1987, 1989, 1990). Furthermore, monkeys regularly fail to acquire fear toward artificial flowers or a toy rabbit using the same paradigm (but they may acquire fear to stimuli such as toy crocodiles which more directly resemble snakes, Cook & Mineka, 1989). These differential acquisition effects could not have been due to either qualitative or quantitative differences in the reactions to FR and FI stimuli displayed by the demonstrator, since the videos were edited to show the demonstrator exhibiting the identical fear behaviors with each of the stimuli. Since the observer monkeys were laboratory bred and had never seen a snake, crocodile, flower or toy rabbit before, Cook & Mineka (1989) concluded: "it seems highly likely that the difference...in the associability of toy snakes..., versus artificial flowers and toy rabbits, derives from phylogenetic rather than ontogenetic factors" (1989, p456-7).
This well-controlled and executed series of studies has contributed substantially to the acceptance of biological preparedness as an explanation of selective associations, but one imbalance in their design leaves room for an alternative explanation of the findings (see also Heyes, 1994). A perplexing aspect of the primate studies is the almost universal failure of observer monkeys to acquire any differential fear to FI stimuli that have been paired with demonstrators behaving fearfully (Cook & Mineka, 1989, 1990). This is curious given the relative ease with which nonprimates acquire fear to FI stimuli by observational learning (e.g. Del Russo, 1975; Kohn, 1976; Mason & Reidinger, 1982; Lore, Blanc & Suedfeld, 1971). Furthermore, if observational learning of fears is as important as some theorists claim (e.g. Rachman, 1977; Mineka, 1987), then it is surprising that these primate studies rarely indicate even a modicum of differential fear to FI stimuli, and all the more so given that these are all novel to the observers (thus ruling out latent inhibition as an explanation of this failure). Differences in salience or discriminability between flower and snake stimuli cannot account for these differences in fear learning because rhesus can solve a complex appetitive discrimination problem at comparable rates regardless of whether the discriminative stimulus is a snake or a flower (Cook & Mineka, 1990; but see Heyes, 1994). It might be argued that the failure to condition fear to FI stimuli caused by the degraded signaling power of video presentation. Even so, there appears to be no qualitative difference in the fear reactions of observers to fear demonstrations presented either in vivo or by video (Cook & Mineka, 1989, p452). Given that at least some clearly distinguishable fear is elicited in the observer by video demonstrations, it is still surprising that no differential fear appears with FI stimuli that have been paired with this reaction.
One radically different explanation of this enigma may be that the UCS used in all of these studies is one that may have relevance solely to snakes. All the fearful reactions in these studies were elicited by a live snake or a very large toy snake; these same reactions were also used to generate observational learning with FI stimuli (flowers and toy rabbits). This is important when we consider that many species of monkey have discriminably different reactions to different predators, some of which are specific to individual predator species such as snakes (cf. Cheney & Seyforth, 1990; see also multiple book review, BBS). For example, from a very early age, infant vervet monkeys are disposed to produce distinctive alarm calls to leopards, eagles and snakes respectively, and this is associated with differential defensive behaviors (in the case of snakes, standing bipedally and peering into the grass around them; Struhsaker, 1967). These alarm calls and defensive posturings appear to convey very specific information about the features and location of the predator concerned (Marler, 1978). Such defensive signals do not appear to be prewired responses to specific stimulus configurations because young vervets learn to associate particular signals with particular predators only after going through a developmental period involving many errors (Cheney & Seyforth, 1990). Hence one possible explanation for failing to find observational conditioning of fear to FI stimuli in the Cook & Mineka studies may be that the fearful demonstrator is signalling specific information about snakes which may be irrelevant to flowers and toy rabbits.
Although there is no direct evidence in the literature that rhesus, like vervets, have a species- specific signal for snakes, they do live in a region (Southern Asia) in which there are many more varieties of snake than elsewhere (Cadle, 1987); thus the conditions would appear optimal for the development of a specific defensive signal for snakes as in the case of vervets. Also, although all the subjects in the Cook & Mineka studies were laboratory bred, this would not have precluded their acquiring knowledge of the signalling system because it is still possible for infant rhesus monkeys to learn representational signalling at a very early age through developmental interactions with adults such as their mothers (Gouzoules, Gouzoules & Marler, 1984; Cheney & Seyforth, 1990). Nor would infant rhesus have to have direct experience with the referent of this signal (snakes) to learn its meaning. For example, the signal could refer to snakes in general or only identify snakes through abstract features (e.g. a snake is a dangerous sinusoidal stimulus). The nature of sinusoidal stimuli could be learned from any stimulus with those properties that the young rhesus encountered. Experiences with certain stimuli that are not snakes (e.g. live insects) can differentially sensitize a fear of snakes in laboratory-bred monkeys (e.g Masataka, 1993). This suggests that experiences with selected nonsnake stimuli can influence subsequent fearful reactions to snakes in some direct or indirect way.
Despite these weaknesses, the studies of observational learning in primates provide the best available evidence for phylogenetically based associations. This evidence would be strengthened if it were found that the rapid learning of fear of snakes in rhesus occurred even when the fear reaction of demonstrators could be shown to be independent of the to- be-conditioned stimulus.
3.4.4 Comparisons of phylogenetic and ontogenetic stimuli
Traditionally, some of the best human laboratory evidence for biological preparedness has come from comparing the conditioning of phylogenetic (e.g. spiders, snakes) and ontogenetic (e.g. weapons, electricity outlets) FR stimuli. The !hman paradigm (see Section 3.4.1) had indicated that resistance to extinction was significantly superior with phylogenetic stimuli than with ontogenetic stimuli (Hugdahl & Karker, 1981; Cook, Hodes & Lang, 1986); this is consistent with predictions from biological preparedness.
Davey (1992a), however, has pointed out that this evidence is incomplete. First, although Hugdahl & Karker (1981) and Cook et al. (1986) both used control conditions with FI stimuli, Cook et al. did not report the results of comparisons between FI and ontogenetic stimuli, and Hugdahl & Karker reported comparisons between these stimuli which were not in a form that allowed the rate of extinction in the two groups to be compared. It is important to compare ontogenetic and FI stimuli: biological preparedness would predict no difference because neither would have been subjected to the selection pressures experienced by phylogenetic stimuli.
Other evidence suggests that resistance to extinction with phylogenetic and ontogenetic stimuli is comparable once the orientation and aversive consequences of the ontogenetic FR stimulus are taken into account. Hugdahl & Johnsen (1989) found that resistance to extinction was strongest when either a gun pointing towards the subject (CS1) was paired with a loud noise (UCS1), or when a snake pointing towards the subject (CS2) was paired with electric shock (UCS2); other combinations of these CSs and UCSs, or conditions where the CSs pointed away from the subject, resulted in poorer resistance to extinction. One interpretation is that the resistance to extinction of phylogenetic and ontogenetic FR stimuli may depend on the semiotic similarity between CS and UCS (cf. Hamm, Vaitl & Lang, 1989) rather than the phylogenetic nature of the CS.
3.4.5 Preparedness effects to subthreshold stimuli
!hman and colleagues have compared FR and FI stimuli which are presented below the threshold for conscious recognition (cf. !hman, 1992ab; !hman, 1993). Using a differential conditioning paradigm where one stimulus (CS+) is always followed by shock and a second (CS-) is never followed by shock, !hman also looked at the effect of backward masking with some of the CS presentations. In the backward masking condition, CS duration was 30 msec and presentation was followed immediately by an irrelevant visual mask. Forced choice recognition tasks using this procedure demonstrated that subjects were unable to recognise backward masked stimuli better than chance (!hman & Soares, 1993; !hman, 1992a).
Even though subjects could not consciously recognise backward masked stimuli, differential conditioning still occurred during acquisition - but only with FR stimuli (e.g. angry faces; Esteves, Dimberg, Parra & !hman, 1993; !hman, 1992a). There was no evidence of differential conditioning to FI stimuli. In a slightly different study, !hman & Soares (1993) conducted differential conditioning to FR and FI stimuli without backward masking; then, during extinction, they introduced backward masking for half the subjects. Backward masking abolished differential responding with FI stimuli but differential fear was maintained throughout extinction with FR stimuli.
!hman (1992) and !hman & Soares (1993) concluded that pairing FRs with aversive UCSs results in conditioned associations controlled at the automatic level of information processing. This does not appear to occur with FI stimuli. !hman (1992a) argues that this automatic processing of FR stimuli occurs because they have critical features which are detected by preattentive processing, and that this "effect of fear-relevant stimuli must be assumed to have an evolutionary origin" (1992, p291).
This kind of subthreshold research is still in its infancy, and there exist alternatives to explanations that stress the evolutionary prewiring of associative or attentional dispositions. The FR stimuli so far used in these studies can be conceived as having evolutionary significance for survival; hence it is reasonable to imply that such stimuli pose a degree of threat that should have selection priority for strategic processing. However, they also have characteristics other than their evolutionary significance that might have high priority for preattentive processing. First, they are dangerous or threatening in some respect. If it is this feature that is important, we might also expect to find differential conditioning despite backward masking to dangerous ontogenetic FR stimuli. Second, phylogenetic FR stimuli also tend to be those that are most feared in Western cultures (Kirkpatrick, 1984). It does not necessarily follow that they are the most feared because they can be labelled potentially dangerous; their fearsomeness could just as well have cultural rather than evolutionary origins (see Section 6). If so, then differential conditioning in the backward masking paradigm might be found with any stimulus of which the subject is afraid. There is some evidence for this. !hman & Soares (1994) selected subjects who feared snakes but not spiders and vice versa; they were then shown backward masked presentations of snakes, spiders, flowers and mushrooms. Subjects exhibited significantly greater phasic skin conductance responses to subthreshold presentations of their phobic stimulus. The responses to their nonphobic FR stimuli did not differ significantly from their responses to FI stimuli. This suggests that it may be prior fear of the stimulus rather than its evolutionary significance per se which enhances preattentive processing. In many respects this effect is similar to the selective attention biases found in anxiety research, particularly where preattentive biases have been demonstrated to stimuli of current threat to the individual (cf. MacLeod & Mathews, 1988; Mogg, Bradley, Williams & Mathews, 1993).
There is a study by Soares & !hman (1993), however, that casts doubt on whether prior fear alone can account for the preattentive processing of subliminally presented FR stimuli. In Soares & !hman's procedure, supraliminal conditioning with FR and FI stimuli was followed by their subliminal presentation during extinction. FR (but not FI) stimuli still evoked a significant physiological CR during backward-masked extinction regardless of whether subjects were high or low on prior fear to the FR stimulus. This suggests that prior fear alone may be insufficient to explain the preattentive processing of FR stimuli. It remains to be discovered, however, whether other characteristics of FR stimuli (e.g. dangerousness, similarity to the UCS) facilitate their preattentive processing.
4. COGNITIVE BIASES IN THE PROCESSING OF FEAR-RELEVANT STIMULI
In the traditional Pavlovian conditioning paradigm, selective associations had been measured in terms of the strength and persistence of conditioned responses (e.g. !hman, 1979). More recently, however, the focus has been on the processing of FR stimuli and aversive consequences. A number of cognitive biases appear to account for differences in the strength and persistence of conditioned fear responses.
4.1. Selective associations and semantic similarity between cue and consequence
Hamm, Vaitl & Lang (1989) have demonstrated that angry faces (CS) become selectively associated with human screams (UCS) and that this selective association depends on the subjects' judgments about the semiotic similarity between cue and consequence. Hugdahl & Johnsen (1989) report results which invite a similar interpretation (see section 3.4.4). Lang (1985) has identified some of the dimensions along which cue and consequence similarity might facilitate selective associations; these include (i) the tendency to approach or avoid both CS and UCS (valence), and (ii) the intensity of arousal to both CS and UCS.
The identification of CS-UCS similarity as an important factor underlying selective associations suggests that some assessment of CS and UCS along selected dimensions may facilitate associative learning between FR stimuli and aversive consequences. If this reflects an inherent characteristic of the cognitive systems processing CS-UCS relationships rather than a prewired predisposition, then CS-UCS similarity should also facilitate associations between ontogenetic FR stimuli and aversive consequences.
4.2 Covariation biases
Recent studies have assessed covariation bias in contingency judgments to FR and FI stimuli. For example, subjects with strong expectancies for covariations between two classes of events often overestimate the contingency between them (cf. Alloy & Tabachnik, 1984; Crocker, 1981). A predisposition to overestimate the contingency between FR stimuli and aversive consequences would be consistent with many of the strong conditioning effects. In a covariation assessment study, Tomarken, Mineka & Cook (1989) exposed subjects to slides of FR and FI stimuli that were followed by shock, tone or nothing. Although the relationship between the slides and outcomes was random, subjects consistently overestimated the contingency between slides of FR stimuli and shock.
A covariation bias of the kind found by Tomarken et al. (1989) appears to occur only with phylogenetic, not with ontogenetic FR stimuli. Covariation assessments with ontogenetic FR stimuli (e.g. weapons, electricity outlets) and aversive outcomes are usually relatively accurate (Sutton, Mineka & Tomarken, 1991; de Jong, Merckelbach, Arntz & Nijman, 1992).
Tomarken et al. (1989) found that initial fear of a stimulus was an important determinant of covariation bias with phylogenetic FR stimuli when the number of stimulus- shock pairings was relatively small (33%), but not when it was increased to 50%. De Jong & Merckelbach (1991) reported that both treated and untreated spider phobics overestimated the relationship between slides of spiders and aversive outcomes. This suggested that prior levels of fear were irrelevant to the covariation bias effect. However, a subsequent study by de Jong, Merckelbach, Arntz & Nijman (1992) found that whereas untreated spider phobics exhibited the expected covariation bias, treated ones did not. The important change in this study was the inclusion of a second FR stimulus (weapons) in the procedure. De Jong et al. (1992) argue that the presence of two competing FR stimuli is likely to undermine the covariation bias between spiders and shock after treatment; it presents treated phobics with an alternative threatening stimulus with which to associate aversive consequences.
If prior fear levels are an important contributor to covariation bias, this might help to explain why only phylogenetic and not ontogenetic FR stimuli exhibit it. Although both are examples of threatening stimuli, the phylogenetic FR stimuli (i.e. snakes and spiders) tend to be those that are most feared within Western cultures (cf. Kirkpatrick, 1984). Hence subjects may begin covariation assessment experiments feeling higher levels of fear toward phylogenetic than ontogenetic stimuli. This still begs the question, of course, of why phylogenetic FR stimuli tend to be more feared than ontogenetic ones; this will be discussed in Section 6.
Finally, the question remains as to the critical cognitive processes that generate an a posteriori covariation bias in these studies. There are at least two important possibilities. First, the covariation bias with FR stimuli may be a computational one which causes an overestimation of the aversive consequences of FR stimuli during the procedure. Alternatively, it may be the result of an expectancy bias that exists prior to the commencement of the procedure.
4.3. UCS expectancy biases
A series of experiments by Davey (1992a) showed that there is a pre-experimental UCS-expectancy bias: subjects enter selective association studies with an inflated estimate of the probability of FR stimuli being followed by aversive consequences. In a 'threat' conditioning procedure, subjects were told that they might receive shock following some stimuli but in fact they received none. Subjects began the experiment with a significantly higher expectancy of aversive UCSs following FR stimuli (snake and spider) than FI stimuli (cat and pigeon). This UCS-expectancy bias with FR stimuli dissipated with continued nonreinforcement, but could be reinstated by a single stimulus-UCS pairing. Davey concluded that (i) this phenomenon could not be the result of a computational bias because subjects received no presentations of the UCS on which computations could be based, and that (ii) this bias could explain the main effects found in the traditional human laboratory selective association conditioning procedure (Section 3.4.1). These include the frequent reports of differential CRs to FR stimuli during preconditioning habituation, their resistance to extinction, and (because of the peculiarities of the design of laboratory selective association conditioning studies) the failure to find significantly superior acquisition with FR stimuli.
Other studies show that phylogenetic and ontogenetic FR stimuli both exhibit a UCS-expectancy bias at the outset of the experimental procedure. Honeybourne, Matchett & Davey (1992) used the 'threat' conditioning procedure of Davey (1992a) to investigate pre-experimental UCS biases to phylogenetic stimuli (snakes, spiders) and ontogenetic ones(guns, electricity outlets) stimuli. Both exhibited an identical UCS- expectancy bias, which was significantly greater than that exhibited by neutral FI stimuli (flowers, landscapes). Similarly, in a 'thought' experiment in which subjects were asked to imagine themselves about to take part in a covariation assessment similar to that carried out by Tomarken et al. (1989), McNally & Heatherton (1993) found that they reported similar pre- experimental probability estimates for both ontogenetic and phylogenetic stimuli and significantly overestimated the probability that each would be followed by shock.
4.4 Comparison of findings from covariation assessment and UCS-expectancy studies
There are some striking but consistent dissimilarities between studies which have tested covariation estimates at the outset (a priori UCS- expectancy studies and covariation assessment 'thought' experiments) and at the end of the procedure. There is a posteriori bias only with phylogenetic FR stimuli, not with ontogenetic ones (Sutton, Mineka & Tomarken, 1991; de Jong, Merckelbach, Arntz & Nijman, 1992) and not when fear to a phylogenetic FR stimulus has been eliminated through behavioural treatment prior to the procedure (de Jong, Merckelbach, Arntz & Nijman, 1992). In contrast, in the a priori studies there were biases towards associating both phylogenetic and ontogenetic FR stimuli with aversive outcomes (Davey, 1992a; Honeybourne, Matchett & Davey, 1993; McNally & Heatherton, 1993), and these biases are found regardless of prior levels of fear (Diamond et al.,1994).
At the very least, these discrepancies mean that a posteriori covariation bias cannot be explained solely in terms of a pre-experimental expectancy bias. McNally & Heatherton (1993) have argued that although phylogenetic and ontogenetic FR stimuli exhibit similar UCS expectancy biases at the outset of the procedure, they differ in their sensitivity to disconfirmation with continued nonreinforcement (in 'threat' conditioning procedures) or continued random reinforcement (in covariation assessment procedures).
One critical factor that may modulate the rate of disconfirmation of the a priori expectancy bias is prior fear. For example, both high fear and low fear subjects exhibit a priori expectancy biases (de Jong, 1993; Diamond, Matchett & Davey, 1993), but only high fear subjects exhibit the a posteriori covariation bias (de Jong, Merckelbach, Arntz & Nijman, 1992). Thus, prior fear may make the a priori bias resistant to disconfirmation. Since phylogenetic FR stimuli also tend to be associated with greater prior fear, this would also explain why both kinds of stimuli exhibit similar a priori expectancy biases but only the phylogenetic ones exhibit covariation bias when measured at the end of the procedure.
5. AN EXPECTANCY BIAS MODEL OF SELECTIVE ASSOCIATIONS
5.1 Expectancy bias: a proximate explanation of selective associations
Having identified an a priori UCS expectancy bias and an a posteriori covariation bias as significant features of selective associations, one must determine their source. As argued, these two biases may not be independent; the a posteriori covariation bias may be determined by the rate of disconfirmation of the a priori expectancy bias (see also McNally & Heatherton, 1993). If so, then identifying the factors that determine (i) the a priori expectancy bias and (ii) the rate of disconfirmation of this bias should provide a predictive model of selective associations.
5.1.1 Determinants of the a priori expectancy bias
The initial expectancy bias is one which can be found with FR or threatening stimuli in general, whether they are phylogenetic or ontogenetic (Honeybourne et al., 1993; McNally & Heatherton, 1993) and whether or not prior fear has accrued to them (Diamond et al., 1993; de Jong, 1993). This suggests that the a priori expectancy bias reflects some initial general assessment of the danger or threat-potential of the stimulus. A number of factors may figure in this initial assessment.
First, there is evidence that an estimate of the danger of the stimulus is a significant predictor of the fear towards it, and hence of the emergence of a UCS- expectancy bias. Merckelbach, van den Hout, Jansen & van der Molen's (1988) subjects assessed a variety of FR and FI stimuli for fearsomeness and danger: a panel of fifteen biologists rated these same stimuli for their relevance to evolutionary survival. There was a highly significant correlation between rated fearsomeness and danger which was only marginally affected when survival relevance was partialled out. In contrast, partialling out rated danger significantly reduced the correlation between fearsomeness and survival relevance. Danger hence seems to be a better predictor of fearsomeness than survival relevance, and survival relevance does not explain the high correlation between fear and danger.
Second, whereas rating the danger of a cue requires assessing the probability of the potential aversive or harmful consequences of encountering that cue in general, other factors which relate to specific features of those consequences may also be important in determining an associative bias. For example, Hamm, Vaitl & Lang's (1989) study suggests that semantic similarity between cue and consequence may be an important factor in selective associations; this is consistent with the finding that preparedness effects occur only when stimuli are paired with certain semantically relevant consequences (e.g. Cook et al., 1986; Hugdahl & Johnsen, 1989; !hman, Frederikson & Hugdahl, 1978). Thus, a UCS- expectancy bias might be expected not only if a stimulus is labelled dangerous, but also if there is semantic covariation between the stimulus and its consequences. Lang (1985) has identified some of the dimensions along which cue and consequence similarity might facilitate selective associations (and, hence, a UCS expectancy bias). These include (i) the tendency to approach or avoid (valence) and (ii) the intensity of activation (arousal). There may be other important dimensions yet to be identified (e.g., similarity of affective response, cf. Greenwald, Cook & Lang, 1989).
Third, it seems intuitively reasonable that prior fear will also inform estimates of the danger of a stimulus. Evidence comes from a study by Diamond et al. (1993). In a factorial design they presented spider phobics and nonphobics with slides of either spiders or kittens. UCS expectancy was measured using the 'threat' conditioning procedure adopted by Davey (1992a). While there was a main effect for spiders at the outset of the procedure (i.e. subjects shown slides of spiders gave higher UCS expectancy ratings regardless of whether they were phobic) there was also a significant picture/phobia interaction in which phobic subjects receiving spider slides gave higher UCS expectancy ratings than nonphobics receiving spider slides. So, while spider slides resulted in higher expectancy ratings in all subjects, these were further increased if the subjects exhibited prior fear to the stimulus.
5.1.2 Determinants of the rate of disconfirmation of a priori expectancy biases
Whereas a priori expectancy biases are found with both phylogenetic and ontogenetic FR stimuli, a posteriori covariation biases are found only with phylogenetic stimuli (Sutton, Mineka & Tomarken, 1991), and then only when subjects start the procedure with relatively high levels of fear (de Jong, Merckelbach, Arntz & Nijman, 1992). These contrasting findings suggest that prior fear may play a role in determining the rate of disconfirmation of a priori expectancy biases.
More direct evidence comes from the study by Diamond et al. (1993) which investigated the effect of prior fear on UCS-expectancy ratings in the 'threat' conditioning procedure. In stage 1 of the procedure (trials 1-3) there was a significant effect of fear- relevance which suggested that higher UCS expectancy ratings were associated with slides of spiders even when the level of prior fear was not considered. However, by stage 2 of the procedure (trials 4-9) this main effect of fear-relevance had disappeared and only a significant interaction between prior fear and fear-relevance remained. This manifested itself as significantly higher expectancies in spider phobic subjects who viewed slides of spiders.
If prior fear does turn out to be the significant factor determining the resistance to disconfirmation of prior expectancies, it could still be argued that prior fear might result from biologically prepared associations and that this is how prepared associations influence both the distribution of fears and their resistance to extinction. However, the restricted range of the so- called prepared stimuli generally used in selective association studies also includes those stimuli which are highly feared within Western cultures (e.g. snakes, spiders, cf. Kirkpatrick, 1984). It does not necessarily follow that these are the most feared because they are biologically prepared; they may be highly feared for cultural reasons. This line of argument will be described in the next section.
5.1.3 Summary of the Expectancy Bias model
Expectancy biases of the kind outlined above do not presuppose specific prewired associations. Such a mechanism would be flexible enough to accommodate both phylogenetic and ontogenetic stimuli as long as (1) there is information available for making judgments about whether they are dangerous and (2) their meaning is clearly defined and understood. It would permit rapid learning about relatively new 'dangerous' stimuli, and could accommodate the social transmission of this information between individuals within their lifetimes. This expectancy bias during the processing of threatening material parallels the attentional biases that have already been identified as a feature of fear- and anxiety-based phenomena (e.g. Mathews & MacLoed, 1985; Mathews, 1990) and would endow the individual with an adaptive benefit not only for attending to and learning about stimuli that were once relevant to our pretechnological ancestors, but also for any stimulus the individual may have reason (through experience or social learning) to label as 'dangerous'.
5.2 Ultimate explanations of expectancy biases
According to the expectancy bias model, selective associations are not simply the result of phylogenetically based associations with "prepared" stimuli, and as such this account differs from the mechanism of selective associations outlined by biological preparedness (Seligman, 1971). However, biological preparedness is primarily an ultimate explanation in that it also attempts to specify the selection pressures that determined which stimuli would exhibit selective associations. What does the approach outlined in this target article have to say about this ultimate explanation?
Expectancy bias permits rapid learning about any stimulus (phylogenetic or ontogenetic) that the individual has reason to suppose is dangerous, so how might a mechanism with such generalized capabilities have been selected for? Biological preparedness supposes that prewired associative predispositions were selected for separately on many different occasions in order to ensure rapid learning about a range of threats. A process that generates biases across a range of threatening stimuli, however, could have resulted from a single evolutionary episode involving a single selection pressure. Since selection does not take place on the basis of a nonfunctional assessment of the utility of the evolved process, even mechanisms with broad adaptive capabilities would probably have been selected for in response to one specific selection pressure (cf. Plotkin, 1983). A nonspecific expectancy bias would have had benefits beyond those associated with the original selection pressure. Such systems are said to contain 'ecologically surplus abilities' (Johnston, 1985), enabling the organism to deal not only with a range of contemporary defensive selection pressures, but, because of the nonspecialized nature of the mechanism, even with sudden catastrophic environmental changes which might introduce a variety of novel and unexpected defensive selection pressures. The prewired associations to specific stimulus configurations called for by biological preparedness clearly would not permit such generalized adaptivity.
Nevertheless, evolution often selects for backup systems for any one function, and arguing that preparedness phenomena can be explained in terms of expectancy bias does not mean that other, more specific associative predispositions may not also have been selected for. Hence the expectancy biases outlined in this paper could also co-exist with more specific prepared associations.
6. CULTURAL INFLUENCES ON SELECTIVE ASSOCIATIONS
According to the expectancy model, judgments about the danger of a stimulus and its semantic and affective relationship to its expected consequences influence selective associations. Hence, to the degree that ontogenetic and cultural factors influence these two types of judgment, these too will influence selective associations. Expectancy bias would predict significant cross-cultural differences in the stimuli that have become fear-inducing, and these should be traceable to the historical events that shaped them (cf. Davey, 1994, for an example of the possible role of cultural factors in the development of fear of spiders in Western cultures).
6.1 Cross-cultural factors determining fear-relevance
Some of the cultural factors that might influence acquired fears include the following: (i) Cultures may differ in the way they transmit information about relatively universal dangers (e.g. information about disease or contamination may influence food taboos, animal phobias and obsessive-compulsive disorders; Simoons, 1961, 1974; Soler, 1979; Tabiah, 1969; Ware, Jain & Davey, 1994). (ii) There can be variation in social taboos or ritualistic behavior patterns based largely on historical dangers which are no longer applicable (e.g xenophobic reactions, sexual fears and phobias). (iii) Information about dangers may be specific to the locality and ecology of the culture (e.g. fear of endemic predatory or venomous animals). (iv) Cultural factors may influence the perception of the causes of different fears (e.g. phobias characterised by somatic complaints are prevalent in many African societies where witch doctors and sorcerers are seen as responsible for abnormal behavior; Morakinyo, 1985; Morakinyo & Akiwowo, 1981). (v) The social acceptability of different phobic symptoms and avoidance responses may vary (e.g. cases of agoraphobia are much rarer in India than in the West, perhaps because of the social acceptability of Indian women being housebound; Chambers, Yergani & Keshavan, 1986; Raguram & Bhide, 1985).
6.2 Problems in the interpretation of cultural influences
Cross-cultural evidence for expectancy bias does not rule out evolved associative predispositions. Cultural factors may have combined with biological predispositions to determine which are the most prevalent FR stimuli in a particular culture. For example, the disproportionate fear of snakes and spiders in Western cultures may have arisen because historical and cultural influences facilitated an existing biological predisposition (see Davey, 1994). Similarly, the fact that some cultures fail to show special fear for so- called "prepared stimuli" (e.g. snakes, spiders) does not necessarily invalidate biological preparedness because there may have been "immunization" against learning the fear as a result of prior exposure to nonthreatening models within the culture (Mineka & Cook, 1986). If phylogenetic predispositions do exist, however, they should be an inherited feature of the human gene pool and this should be experimentally detectable across cultures as a significant resistance to extinction in the laboratory differential conditioning paradigm (see Section 3.4.1).
Finally, it has been taken as support for biological preparedness that in Western cultures the stimuli that are feared the most are those which are describable as phylogenetic rather than ontogenetic FR stimuli (see Section 1). However, phylogenetic FR stimuli have by definition been around significantly longer than ontogenetic FR stimuli; hence any fear- relevance would have had more time to become established through absorption into cultural traditions and values (cf. Ingold, 1988; Willis, 1990; Nichter, 1981; Renner, 1990; Mundkur, 1983). The typical ontogenetic FR stimulus (e.g. guns, electricity outlets) have not existed long enough to acquire the kind of symbolic significance that would perpetuate their fear relevance.
6.3 The evolutionary significance of cultural transmission
There are important adaptive benefits to the cultural transmission of fears. It tends to be forgotten that cultural transmission is an exceptionally efficient method of transfering information about threats to survival from one generation to the next and it is likely to become a prime source of adaptation when phylogenesis has reached an upper limit to the amount of change it can cope with in a relatively variant environment (cf. Plotkin & Odling-Smee, 1979, 1981; Ruyle, 1973; Durham, 1976). This alone would seem to provide a good theoretical reason for exploring more fully the possible influence of culture on fear distribution. In combination with expectancy biases, it suggests a viable alternative to genetic transmission of specific fears.
7. FINAL SUMMARY
This target article has reviewed the evidence supporting biological and cognitive explanations of selective associations. It has been argued that a bias in the processing of information about threatening stimuli causes a heightened expectation of aversive outcomes following FR stimuli which in turn generates and maintains a robust learned association between them. Some of the features of FR stimuli which determine this expectancy bias are estimates of their danger, the semiotic similarity between them and their aversive outcomes, and the degree of prior fear they elicit. Since ontogenetic and cultural factors also influence these features of FR stimuli, they will play an important role in determining expectancy biases. The available evidence does not exclude the possibility that both expectancy biases and specific evolved predispositions may co-exist, but expectancy bias can account for most of the important findings that cannot be attributed to specific evolved predispositions.
Agras W.S., Sylvester D. & Oliveau D. (1969) The epidemiology of common fears dphobias.ComprehensivePsychiatry,10,151-156.
Alloy L.B. & Tabachnik N. (1984) Assessment of covariation by humans and animals: The joint influence of prior expectations and current situational information. Psychological Review, 91, 149.
Bennett-Levy J. & Marteau T. (1984) Fear of animals: What is prepared? British Journal of Psychology, 75, 35-42.
Cadle J.E. (1987) Geographic distribution: Problems in phylogeny and zoogeography. In Seigel R.A., Collins J.T. & Novak S.S. (Eds) Snakes, ecology and evolutionary biology, New York: Macmillan Publishing Co.
Chambers J., Yeragani V.K. & Kershavan M.S. (1986) Phobias in India and the United Kingdom: A trans- cultural study. Acta Psychiatrica Scandinavia, 74, 388-391.
Cheney D.L. & Seyforth R.M. (1990) How monkeys see the world, Chicago: Chicago University Press.
Cook E.W. III, Hodes R.L. & Lang P.J. (1986) Preparedness and phobias: Effects of stimulus content on human visceral conditioning. Journal of Abnormal Psychology, 95, 195-207.
Cook M. & Mineka S. (1987) Second order conditioning and overshadowing in the observational conditioning of fear in monkeys. Behaviour Research & Therapy, 25, 349- 364.
Cook S. & Mineka S. (1989) Observational conditioning of fear to fear-relevant versus fear-irrelevant stimuli in rhesus monkeys. Journal of Abnormal Psychology, 98, 448-459.
Cook M. & Mineka S. (1990) Selective associations in the observational conditioning of fear in rhesus monkeys. Journal of Experimental Psychology: Animal Behavior Processes, 16, 372-389.
Cook M., Mineka S., Wolkenstein B. & Laitsch K. (1985) Observational conditioning of snake fear in unrelated rhesus monkeys. Journal of Abnormal Psychology, 94, 591-610.
Costello C.G. (1982) Fears and phobias in women: A community study. Journal of Abnormal Psychology, 91, 280-286.
Crocker J. (1981) Judgment of covariation by social perceivers. Psychological Bulletin, 90, 272-292.
Davey G.C.L. (1992) An expectancy model of laboratory preparedness effects. Journal of Experimental Psychology: General, 121, 24-40.
Davey G.C.L. (1994) The 'disgusting' spider: The role of disease and illness in the perpetuation of fear of spiders. Society & Animals, 2, 17-24.
De Jong P. (1993) Covariation bias in phobia: Mere resistance to preexperimental expectancies? Behavior Therapy, 24, 447-454.
De Jong P. & Merckelbach H. (1991) Covariation bias and electrodermal responding in spider phobics before and after behavioural treatment. Behaviour Research & Therapy, 29, 307-314.
De Jong P., Merckelbach H., Arntz A. & Nijman H. (1992) Covariation detection in treated and untreated spider phobics. Journal of Abnormal Psychology, 101, 724-727.
Del Russo J.E. (1975) Observational learning of discriminative avoidance in hooded rats. Animal Learning & Behavior, 3, 76-80.
de Silva P. (1988) Phobias and preparedness: Replication and extension. Behaviour Research & Therapy, 26, 97-98.
de Silva P., Rachman S. & Seligman M.E.P. (1977) Prepared phobias and obsessions: Therapeutic outcome. Behaviour Research & Therapy, 15, 65-77.
Diamond D., Matchett G. & Davey G.C.L. (1993) The effect of prior fear levels on UCS-expectancy ratings to a fear-relevant stimulus. Quarterly Journal of Experimental Psychology, A, in press.
Durham W.H. (1976) The adaptive significance of cultural behavior. Human Ecology, 4, 89-121.
Esteves F., Parra C., Dimberg U. & !hman A. (1993) Nonconscious associative learning: Pavlovian conditioning of skin conductance responses to masked fear-relevant facial stimuli. Psychophysiology, in press.
Gouzoules S.H., Gouzoules H. & Marler P. (1984) Rhesus monkey (Macac mulatta) screams: Representational signalling in the recruitment of agonistic aid. Animal Behaviour, 32, 182-193.
Greenwald M.K., Cook E.W. & Lang P.J. (1989) Affective judgment and psychophysical response: Dimensional covariation in the evaluation of pictorial stimuli. Journal of Psychophysiology, 3, 51-64.
Hamm A.O., Vaitl D. & Lang P.J. (1989) Fear conditioning, meaning, and belongingness: A selective analysis. Journal of Abnormal Psychology, 98, 395-406.
Heyes C.M. (1994) Social learning in animals: Categories and mechanisms. Biological Review, in press.
Honeybourne C., Matchett G. & Davey G.C.L. (1993) An expectancy model of preparedness effects: A UCS- expectancy bias in phylogenetic and ontogenetic fear- relevant stimuli. Behavior Therapy, 24, 253-264.
Hugdahl K. & Johnsen B.H. (1989) Preparedness and electrodermal fear-conditioning: Ontogenetic vs. phylogenetic explanations. Behaviour Research & Therapy, 27, 345-353.
Hugdahl K. & Karker A.C. (1981) Biological vs. experimental factors in phobic conditioning. Behaviour Research & therapy, 19, 109-115.
Ingold T. (1988) What is an animal? Unwin Hyman: London.
Johnston T.D. (1985) Introduction: Conceptual issues in the ecological study of learning. In T.D. Johnston & A.T. Pietrewicz (Eds) Issues in the ecological study of learning. Hillsdale, N.J.: Erlbaum.
Kirkpatrick D.R. (1984) Age, gender and patterns of common intense fears among adults. Behaviour Research & Therapy, 22, 141-150.
Kohn B. (1976) Observational and discrimination learning in the rat: Effects of stimulus substitution. Learning & Motivation, 7, 303-312.
Lang P.J. (1985) The cognitive psychophysiology of emotion: Fear and anxiety. In A.H. Tuma & D. Maser (Eds) Anxiety and the anxiety disorders. Hillsdale, N.J.: Erlbaum.
LoLordo V.M. & Droungas A. (1989) Selective associations and adaptive specializations: Taste aversions and phobias. In S.Klein & P.Mowrer (Eds) Contemporary learning theories: Instrumental conditioning theory and the impact of biological constraints on learning. Hillsdale, N.J.: Erlbaum.
Lore R., Blanc A. & Suedfeld P. (1971) Empathetic learning of a passive-avoidance response in domesticated Rattus Norvegicus. Animal Behaviour, 19, 112-114.
MacLeod C. & Mathews A. (1988) Anxiety and the allocation of attention to threat. Quarterly Journal of Experimental Psychology, 40, 653-670.
Marler P. (1978) Affective and symbolic meaning: some zoosemiotic speculations. In T.A. Sebeok (Ed) Sight, sound and sense. Bloomington, Ind.: Indiana University Press.
Masataka N. (1993) Effects of experience with live insects on the development of fear of snakes in squirrel monkeys (Saimiri sciureus). Animal Behaviour, 46, 741-746.
Mason J.R. & Reidinger R.F. (1982) Observational learning of food aversions in red-winged blackbirds (Agelaius phoeniceus). Auk, 9, 548-554.
Mathews A. (1990) Why worry? The cognitive function of anxiety. Behaviour Research & Therapy, 28, 455-468.
Mathews & McLeod C. (1985) Selective processing of threat cues in anxiety states. Behaviour Research & Therapy, 23, 563-569.
McNally R.J. (1987) Preparedness and phobias: A review. Psychological Bulletin, 101, 283-303.
McNally R.J. & Heatherton T.F. (1993) Are covariation biases attributable to a priori expectancy biases? Behaviour Research & Therapy, in press.
Merckelbach H., van den Hout M.A., Hoekstra R. & Van Oppen P. (1988) Many stimuli are frightening but some are more frightening than others: The contribution of preparedness, dangerousness, and unpredictability to making a stimulus fearful. Journal of Psychopathology and Behavioural Assessment, 10, 355-366.
Mineka S. (1985) Animal models of anxiety-based disorders. In A. Tuma & J. Maser (Eds) Anxiety and anxiety disorders. Hillsdale, N.J.: Erlbaum.
Mineka S. (1987) A primate model of phobic fears. In H.J. Eysenck & I. Martin (Eds) Theoretical foundations of behaviour therapy. New York: Plenum Press.
Mineka S., Davidson M., Cook M. & Weir R. (1984) Observational conditioning of snake fear in rhesus monkeys. Journal of Abnormal Psychology, 93, 355- 372.
Mogg K., Bradley B.P., Williams R. & Mathews A. (1993) Subliminal processing of emotional information in anxiety and depression. Journal of Abnormal Psychology, 102, 304-311.
Morakinyo O. (1985) Phobic states presenting as somatic complaintes in Nigeria: Socio-cultural factors associated with diagnosis and psychotherapy. Acta Psychiatrica Scandinavia, 71, 356-365.
Morakinyo O. & Akiwowo A. (1981) The Yoruba ontology of personality and motivation. A multidisciplinary study. Journal of Social and Biological Structures, 4, 19-38.
Mundkur B. (1983) The cult of the serpent. Albany: State University of New York Press.
Nichter M. (1981) Idioms of distress: Alternatives in the expression of psychosocial distress - a case study from South India. Culture, Medicine & Psychiatry, 5, 379-408.
!hman A. (1979) Fear relevance, autonomic conditioning, and phobias: A laboratory model. In P.O. Sjoden, S. Bates & W.S. Dockens III (Eds) Trends in behavior therapy, New York: Academic Press.
!hman A. (1992a) Orienting and attention: Preferred preattentive processing of potentially phobic stimuli. In A. Campbell, H. Hayne & R. Richardson (Eds) Attention and information processing in infants and adults: Perspectives from human and animal research. Hillsdale, N.J.: Erlbaum.
!hman A. (1992b) Stimulus prepotency and fear: data and theory. In N. Birbaumer & A. !hman (Eds) The organization of emotion: Cognitive, clinical and psychophysiological perspectives. Toronto: Hogrefe.
!hman A. (1993) Fear and anxiety as emotional phenomena: Clinical phenomenology, evolutionary perspectives, and information processing mechanisms. In M. Lewis & J.M. Haviland (Eds) Handbook of emotions. New York: Guilford Publications.
!hman A., Dimberg U. & !st L.G. (1985) Animals and social phobias: Biological constraints on learned fear responses. In S. Reiss & R.R. Bootzin (Eds) Theoretical issues in behavior therapy. New York: Academic Press.
!hman A., Fredrikson M.L. & Hugdahl K. (1978) Orienting and defensive responding in the electrodermal system: Palmar-dorsal differences and recovery rate during conditioning to potentially phobic stimuli. Psychophysiology, 15, 93-101.
!hman A. & Soares J.J.F. (1993) On the automatic nature of phobic fear: Conditioned electrodermal responses to masked fear-relevant stimuli. Journal of Abnormal Psychology, 102, 121-132.
!hman A. & Soares J.J.F. (1994) "Unconscious anxiety": Phobic responses to masked stimuli. Journal of Abnormal Psychology, in press.
Plotkin H.C. (1983) The functions of learning and cross- species comparisons. In G.C.L. Davey (Ed) Animal models of human behaviour. Chichester: Wiley.
Plotkin H.C. & Odling-Smee F.J. (1979) Learning, change and evolution. Advances in the study of behavior, 10, 1- 41.
Plotkin H.C. & Odling-Smee F.J. (1981) A multiple-level model of evolution and its implications for sociobiology. The Behavioral & Brain Sciences, 4, 225- 268.
Plotkin H.C. & Odling-Smee F.J. (1982) Learning in the context of a hierarchy of knowledge gaining processes. In H.C. Plotkin (Ed) Essays in evolutionary epistemology. Chichester: Wiley.
Rachman S. (1977) The conditioning theory of fear- acquisition: A critical examination. Behaviour Research & Therapy, 15, 375-387.
Raguram R. & Bide A.V. (1985) Patterns of phobic neurosis: A retrospective study. British Journal of Psychiatry, 147, 557-560.
Renner F. (1990) Spinnen: Ungeheuer - sympathisch. Verlag: Kaiserslautern.
Russell P.A. (1979) Fear evoking stimuli. In W. Sluckin (Ed) Fear in animals and man. New York: Van Nostrand Reinhold.
Ruyle E.E. (1973) Genetic and cultural pools: Some suggestions for a unified theory of biocultural evoluation. Human Ecology, 1, 201-215.
Seligman M.E.P. (1970) On the generality of the laws of learning. Psychological review, 77, 406-418.
Seligman M.E.P. (1971) Phobias and preparedness. Behavior Therapy, 2, 307-320.
Simoons F.J. (1961) Eat not this flesh. Madison: University of Wisconsin Press.
Simoons F.J. (1974) Fish as forbidden food: The case of India. Ecology of food and nutrition, 3, 185-201.
Soares J.J.F. & Ohman A. (1993) Backward masking and skin conductance responses after conditioning to nonfeared but fear-relevant stimuli in fearful subjects. Psychophysiology, 30, 460-466.
Soler J. (1979) The semiotics of food in the Bible. In R. Forster & O. Randum (Eds) Food and drink in history. Selections from Annals. Volume 5. Baltimore: John Hopkins University Press.
Struhsaker T.T. (1967) Auditory communications among vervet monkeys. (Cercoopithecus aethiops). In S.A. Altman (Ed) Social communication among primates. Chicago: Chicago University Press.
Sutton S.K., Mineka S. & Tomarken A.J. (1991) Affective versus semantic determinants of covariation bias between fear-relevant stimuli and aversive outcomes. Paper presented at the meeting of the Midwestern Psychological Association, Chicago, Ill.
Tambiah S.J. (1963) Animals are good to think and good to prohibit. Ethnology, 8, 423-459.
Tomarken A.J., Mineka S. & Cook M. (1989) Fear-relevant selective associations and covariation bias. Journal of Abnormal Psychology, 98, 381-394.
Ware J., Jain K., Burgess I. & Davey G.C.L. (1994) Desease- avoidance model: factor analysis of common animal fears. Behaviour Research & Therapy, 32, 57-63.
Willis R.G. (1990) Signifying animals. Unwin Hyman: London.
Zafiropoulou M. & McPherson F.M. (1986) "Preparedness" and the severity and outcome of clinical phobias. Behaviour Research & Therpay, 24, 221-222.
The author is grateful to Peter de Jong, Rich McNally and Harold Merckelbach for their comments on earlier versions of this paper. It was written while the author was supported by project grant R000234340 from the Economic & Social Research Council, U.K.