Negativity bias: An evolutionary hypothesis and an empirical programme. John Lazarus. Learning and Motivation, Volume 75, August 2021, 101731. https://doi.org/10.1016/j.lmot.2021.101731
Highlights
• Hypothesis: negativity bias evolves when fitness is a concave function of state.
• Threat explanation of negativity bias unsound if based on incommensurate stimuli.
• Incommensurate stimuli can be studied for bias using the loss aversion paradigm.
• The ‘potency equivalence function’ measures equipotency of incommensurate stimuli.
Abstract: Across many psychological domains there is evidence of negativity bias: the greater subjective potency of negative events when compared with positive events of the same objective magnitude. Here I propose a general evolutionary explanation for the phenomenon: the concave fitness-state (CFS) hypothesis. The CFS hypothesis proposes, with evidence from feeding, drinking and economic domains, that various motivational, emotional and cognitive states – through which stimuli activate responses – have a concave downwards (diminishing returns) relationship with fitness. Where this is the case it follows that <. In discussing other approaches to understanding the phenomenon I critique the proposal that negativity bias can be explained as an adaptive response to the particular importance and urgency of dealing with threat, by arguing that: (1) where negative stimuli interpretable as threat, and contrasting positive stimuli, cannot be measured in a commensurate manner they cannot be validly tested for negativity bias; and (2) since threat stimuli and positive stimuli generally impact different states a greater potency for threat stimuli should generally be interpreted in terms of motivational competition rather than negativity bias. I suggest two ways of circumventing the problem of incommensurate stimuli when studying stimulus bias. The first is to use the loss aversion paradigm: rating the value of the same stimulus when presented as either a gain or a loss in relation to a reference value. Second, understanding the relative subjective potencies of positive and negative stimuli across a range of objective stimulus magnitudes, even when incommensurate, can be achieved experimentally by finding pairs of positive and negative stimuli which, though measured on different scales of magnitude, are equipotent. That is, they have equal and opposite effects on fitness, well-being or stimulus evaluation. These stimulus pairs constitute a potency equivalence function, which describes the shape of the relationship between equipotent positive and negative stimulus magnitudes.
2. Current evolutionary explanations for negativity bias
2.1. Threat
While a number of causal mechanisms have been proposed for the negativity bias phenomenon (Cacioppo, Gardner, & Berntson, 1997; Baumeister et al., 2001; Cacioppo & Berntson, 1994; Kellermann, 1984; Park & Van Leeuwen, 2014; Rozin & Royzman, 2001; Taylor, 1991) there are rather few evolutionary explanations, the dominant idea focussing on threat:
‘To the extent that it is more difficult to reverse the consequences of an injurious or fatal assault than an opportunity unpursued, a propensity to react more strongly to negative than positive stimuli may have developed through the process of natural selection’ (Cacioppo & Berntson, 1994: 413, emphasis added).
‘From our perspective, it is evolutionarily adaptive for bad to be stronger than good. We believe that throughout our evolutionary history, organisms that were better attuned to bad things would have been more likely to survive threats and, consequently, would have increased probability of passing along their genes. As an example, consider the implications of foregoing options or ignoring certain possible outcomes. A person who ignores the possibility of a positive outcome may later experience significant regret at having missed an opportunity for pleasure or advancement, but nothing directly terrible is likely to result. In contrast, a person who ignores danger (the possibility of a bad outcome) even once may end up maimed or dead. Survival requires urgent attention to possible bad outcomes, but it is less urgent with regard to good ones. Hence, it would be adaptive to be psychologically designed to respond to bad more strongly than good.. . . At the broadest level, we argue that bad is stronger than good because responding to the world in this way is adaptive.. . . This argument is admittedly speculative.’ (Baumeister et al., 2001: 325, 357, emphasis added).
‘In the extreme, negative events are more threatening than are positive events beneficial. The clear example here is death, a final, irreversible event. Avoiding risks of death must be a matter of the highest priority in the evolutionary scheme; the peak of vigilance and investment would well be oriented to escape death. It is true that reproduction is the final measure of evolutionary success, but there are usually multiple opportunities to reproduce, and death terminates these options.. . . Negative events often develop more rapidly and require a rapid response. The model, of course, is predator threat.. . . Negative events. . . require a more sophisticated appraisal, because the options for action [related to threat] are more varied [than for]. . . positive entities’ (Rozin & Royzman, 2001: 314).
‘When directly compared or weighted against each other, losses loom larger than gains. This asymmetry between the power of positive and negative expectations or experiences has an evolutionary history. Organisms that treat threats as more urgent than opportunities have a better chance to survive and reproduce’ (Kahneman, 2012: 282).
Others have expressed the same view, including the need for an urgent response to threat stimuli (Norris, 2019: 3; Peeters & Czapinski, 1990: 55; Pratto & John, 1991: 380; Taylor, 1991: 78).
Considered broadly, the impact of threat can be overestimated. In particular, prey animals maintain a level of vigilance to pre-empt the risk of undetected attack and adjust this level in response to changes in predation risk and competing demands (Beauchamp, 2015). Responses to a change in threat level can therefore be subtle and involve little cost. For example, risk increases as group size declines and birds adjust vigilance level accordingly, increasing and decreasing it in response to the stimulus of individuals leaving the group and joining it, respectively (Roberts, 1995). (This also exemplifies the broader point that a response to one stimulus has to be understood in the context of conflicting stimuli and internal states: section 4.2.) Further, the need for urgency emphasised in the above quotes is also crucial for a successful response to some positive events, such as a prey item that needs to be chased and captured, a potential mate that must be courted, or any rewarding stimulus subject to competition with others, and Kahneman (2012: 301) makes a similar point. All this being said, however, and as the emphases in the first two quotes above argue, threat does have a particular potency in that failure to respond efficiently to an imminent attack can reduce fitness greatly and irrecoverably, or be fatal.
But how relevant for the concept of negativity bias is this contrast between threat and the qualitatively different positive events and opportunities of the above quotes that represent what I will call the ‘threat hypothesis’? The positive and negative events here – an opportunity to mate or obtain a resource versus the risk of assault, say – are, it seems, largely both stimulus-incommensurate and, more importantly, state-incompatible. And if the relevant stimuli are stimulus-incommensurate it is difficult to see how their magnitudes can be validly compared using existing methods in order to test for negativity bias (see section 1.2) and thus to test the threat hypothesis itself; in sections 5.2 and 5.3 I suggest techniques that can overcome this problem. In addition, there is a particular problem with the threat hypothesis if used to explain the evolution of choices between actions when state-incompatible positive and negative stimuli are present simultaneously. In this case it seems more valid to explain the outcome in terms of competition between different states for the control of stimulus evaluation and behaviour rather than in terms of stimulus bias and without reference to state (and I use ‘competition’ informally here, rather than in any technical motivational sense; e.g. McFarland, 1974).
This argument holds most clearly for apparently natural behaviours in the real world, the kinds of behaviour imagined in the above quotes arguing for an evolutionary role for threat in the negativity bias phenomenon. But what happens, conceptually, when we bring participants from the real world into the lab and linguistic and visual stimuli carry negative meanings which are measurable in ways that can be shared with positive stimuli? Since this does not in itself make the positive and negative stimuli state-compatible, I would argue that it is not sufficient to validate a negativity bias analysis. Again, the relative potency of different states, rather than of stimuli acting on the same state, is at issue.
And what if the negative attributions in such a lab study were processed by neural systems evolved to deal with threats to the person? Does this change the argument and how far might the threat hypothesis go then in explaining negativity bias across the very disparate domains in which it has been suggested to have a role? Although many negativity bias experiments employ negative stimuli that are not obviously physically threatening some studies seem to suggest that neural processes classify such stimuli as fear inducing. The speed of processing of disagreeable ethical statements (Van Berkum, Holleman, Nieuwland, Otten, & Murre, 2009) is one example. Another is activation of the amygdala, which is associated with the processing of negative emotional stimuli, including threat, but also with positive stimuli (Adolphs, Tranel, Damasio, & Damasio, 1995; Toates, 2007: 318–320). Although amygdala activation occurs during loss aversion – a stronger evaluation against a loss than for an equivalent gain (section 3.3) – I have found evidence for this only for gambles rather than the riskless evaluation typical of the negativity bias literature (De Martino, Camerer, & Adolphs, 2010; Kahn et al., 2002; Sokol-Hessner, Camerer, & Phelps, 2013). Since the amygdala is also responsive to uncertainty without biological relevance (Herry et al., 2007; Hsu, Bhatt, Adolphs, Tranel, & Camerer, 2005) these neural responses to loss aversion in a gambling context may represent heightened vigilance in response to uncertainty (Whalen, 2007) rather than a reaction to threat. And importantly it would be good to know more about neural processing in riskless evaluation more relevant to negativity bias; Garavan, Pendergrass, Ross, Stein, and Risinger (2001) found equal amygdala activation for positive and negative stimuli of roughly matched magnitude.
All this being said, if it turns out that physically innocuous negative stimuli are routinely processed in the same way as threats to the person this still seems to leave these negative stimuli in a qualitatively different and state-incompatible condition to most positive stimuli. And this state-incompatibility is an even stronger reason for concern about the potential over-generalization of the negativity bias phenomenon than that based on stimulus-incommensurateness and pointed out by others (Norris, 2019; Rozin & Royzman, 2001: 300; Taylor, 1991: 68). To repeat, competition seems to be the correct concept when considering the interaction between incompatible states and it would be conceptually preferable to reserve the notion of negativity bias for the phenomenon of bias based on the unequal potency of stimuli of equal objective magnitude.
2.2. Other evolutionary accounts
Park & Van Leeuwen’s (2014: 89–90) asymmetric behavioural homeostasis hypothesis ‘conceptualizes many motivational processes as 1-sided homeostatic mechanisms and. . . predicts that motivational responses. . . amplified by certain cues will not be reversed simply by reversing the input cues. . . [so] that many evolutionarily adaptive. . . responses to fitness threats (e.g., fears, aversions) are more easily inflamed than dampened’.
Rozin and Royzman (2001: 314) point to a class of negative contagious events, ‘[t]he basic model [being] the germ, for which there is not an obvious positive parallel’ and which has ‘by a process of preadaptation, spread through other domains of life (such as morality)’.
Finally, when there is uncertainty about the nature or existence of events, adaptive decisions will take account not only of their likely consequence, good or bad, but also of their likelihood of occurrence. Signal detection theory and error management theory then provide the methods for calculating the potency of stimuli and any cognitive biases that emerge from best responses. Negativity bias may then be predicted for events including threatening stimuli, contaminants and biases in interpersonal perception (Haselton & Nettle, 2006).
2.3. Conclusion
A number of evolutionary accounts of negativity bias, of varying degrees of potential generality, have been provided.
In the following section I propose an evolutionary hypothesis to explain negativity bias which is potentially of wide generality, and in section 3.3 I consider the relationship between this hypothesis and some other accounts of negativity bias, evolutionary, psychological and economic.