Though the discussion thus far has been largely devoted to the complexities that a serious RST would have to grapple with, there are also positive recommendations to be made. But all the various decision points (which model, which traits, which costs/benefits etc) are not independent, and the search space for reasonable applications of signalling theory to religious signalling can be narrowed somewhat. Before concluding with a breakdown of the remaining open questions then, one set of these recommendations I will offer are some tentative ‘templates’ for modelling religious signalling. The idea here is to practice what thus far has only been preached; and try to bundle together some of the various options into plausible, potentially useful ‘templates’ for application.
6.4.1. Cognitive constraint templates (hard to fake)
Beginning with hard-to-fake signals (in the sense of index or constraint signals), we have already seen (via Sterelny) one interpretation where dishonest commitment signals become hard to fake: where lies either elicit too much cognitive dissonance or weave too tangled a web for the talents and efforts of the liar to reliably overcome. This will lean on the Frankian notion discussed earlier: that emotional/affective responses providing a hard-to-fake, honest window onto the true commitments of religious participants.
Consider the following reading. Intense or demanding religious rituals are culturally evolved ways of eliciting commitment-indexing affective responses: either hard-to-fake displays of enthusiasm and engagement from the genuinely committed, or hard-to-avoid giveaways of discomfort and dissonance from those only taking part for instrumental reasons. Taking part in these rituals might incur fitness-relevant costs, but they need not, e.g. experiencing pain during an extreme religious ritual is ‘costly’ in some intuitive sense of the word but might not translate into an evolutionarily relevant cost. Considering the evolutionary pressures on signal costs in such a model, one might therefore expect the evolution of this mode of religious signalling to optimise away from fitness-impacting demands, in favour of discomfort, pageantry, and awe. This Frankian template for cognitive constraint-based religious signalling ticks several boxes. It includes a clear generative link between prosocial commitment and reliable signals. It looks like a good fit with many religious rituals and public worship practices, whether high emotional intensity to elicit enthusiasm or dissonance, or low intensity to probe for boredom and resentment (in Quaker meeting for worship, fidgeting can be a dead giveaway). Extreme, painful rituals would also fit this template, given the apparent cognitive penetration of religious devotion into the experience of such rituals (Jegindø et al. 2013).
But it also has limitations. Revealing cognitive constraints by placing senders in a specific constructed situation is not obviously applicable outside the ritual context. As an index signalling model it does not predict seemingly arbitrary demands with fitness-significant costs; such as direct tithing or sacrifice of resources, or restrictions on dress, diet, economic activity, and sexual practices. And it does not explicitly predict signal complexity or cross-cultural diversity, as it is the emotional responses of the participants that constitute the signal, not the specific tropes and components of the rituals themselves. And any scepticism about the reliability, fakeability, or stability of moral emotional displays would have an impact here, see e.g. (Bandes 2016; Deem and Ramsey 2016). And in the abstract, unfakeable-signalling interpretations need exogenous constraints or connections to leverage, bringing with them both synchronic robustness requirements and the need for a further evolutionary back-story (as with cognitive biases in non-adaptive, non-signalling explanations).
Such considerations are not knock-down objections, and for signalling interpretations of religious ritual this is perhaps the most appropriate template to use, but they highlight gaps which differential cost-benefit mechanisms might instead fill.
6.4.2. Differential cost templates: punishment and bridge-burning
With respect to differential-cost fakeable signalling, we already have one option on the table: Sosis’s hybrid model where the evolution of signalling strategies is driven by subjective perceptions of cost (rather than by objective cost-benefit), and the signalling (pseudo-)system is bootstrapped by and co-evolves with benefits from cooperative assortment.
But suppose we also let signals and responses be temporally extended or distributed, rather than always associated with discrete events like rituals. For example, (as suggested earlier in response to Bulbulia), signals might map to persistent adherence to codes of dress or practice, perhaps where the absence of lapses or exceptions functions as the signal, rather than any one- off adherence event. In this vein, signalling strategies involving costly signals also need not be strictly ‘pay as you go’. A comprehensive survey of all the possibilities will not be attempted, however, there are potential examples worth delving into with respect to temporally extended signalling.
The first comes with punishment of defection, which some argue should be folded into the notion of signal cost (Fraser 2011; Murray and Moore 2009). Suppose that acceptance into a community after signalling commitment now places you at the mercy of that community’s retribution, should you defect from your commitment. In terms of cost-benefit analysis, such a signal of commitment followed by defection would be equivalent to signalling that you’re a high type but being a low type (high types are the ones who don’t defect). This might seem counter-intuitive, but although the order of signal and response is important in modelling sender-receiver reactions, the order of associated cost and benefit is not: they only matter for purposes of determining overall payoffs. This means that the whole temporally extended scenario can be modelled as a single step in an evolutionary signalling game in which both high and low types pay the same (if anything) at the time the commitment signal is sent, but signalling low types also reliably pay an extra cost of being caught out.
Importantly, this is a signalling game with differential costs which meets Henrich’s differential objectivity challenge and avoids the need for Sosis’s retreat to subjective differential costs. The evolutionary proviso is that the overall behaviour of the community and the individual agent must be governed by strategies which are updated or reinforced in a way which approximates an adaptive response with respect to overall objective payoffs. If so then those strategies can co-evolve, and a stable signalling system emerge.
A related template is ‘bridge burning’. Consider the case of group-specific ritual scarification, tattoos, and other permanent markers. One powerful way of promoting group cohesion is parochial altruism (Bowles and Gintis 2011): treating ingroup members well while treating outgroup members badly (or simply excluding them – in our traditional evolutionary context humans are obligate community-members). In such an environment, an agent who permanently marks themselves as belonging to a certain group has made a powerful statement of commitment, because they have ‘burnt their bridges’ and dramatically increased their own cost of defection (Sterelny 2012a). If we again ignore the temporal ordering of cost and benefit and only look at the overall payoffs then it is clear that permanently marking oneself out as exclusive to a particular group is a differentially costly persistent signal: inexpensive for the genuinely committed, dangerous for those likely to leave the community (or fall out with them). But in this case the community need not have a reliable defection-punishment at all, because bad consequences for defection are the result of general outgroup hostility and the signals themselves.
Viewed as temporally extended modelling targets, punitive cultural practices and persistent, group-specific markers like tattoos and ritual scarification can therefore form the moving parts of differentially costly signalling systems. Technically, neither template exactly maps the classic handicap model, as signal cost is vulnerable to community/receiver response (e.g. in ‘punishment’, signalling is only costly for low types who are actually accepted), however in both cases the crucial free-rider game path (being low, signalling high, treated as a high) is made prohibitively costly. More importantly for current purposes, neither interpretation is entirely free-standing in the idealised, maximally parsimonious sense, as they rely on various cultural practices (punishment, parochial altruism) the origins and stability of which call for additional explanation.
6.4.3. Differential benefit and other templates
There is a relatively simple differential cost-benefit template which avoids such complications, and the differential-cost model altogether: differential benefit signalling.
Imagine two individuals considering initiation into a local community, call them Flaky and Staunch. The community is good at productively combining the labours of its members and fairly distributing the benefits (making it attractive to join), but there are significant transaction costs for absorbing new members, and while Staunch is enthusiastically attracted to the group
and its way of doing things, Flaky is far less committed and just looking to take what’s on offer until a better option turns up. For Staunch, there is little prospect of wanting to leave the community or falling afoul of its rules, but Flaky has a reasonable risk of messing up at some point, or just wanting out. So, while Staunch can look forward to staying in the community for life (if admitted), Flaky’s future in it has a comparatively short half-life. This means that the total future benefits for Staunch will be significantly greater than those for Flaky, but the net benefits for the community would also be different: Staunch would be an asset but investing in Flaky is less likely to pay off (for similar reasoning with regard to the evolution of guilt, see (O’Connor 2016; Rosenstock and O’Connor 2018))
This straight-forwardly fits the sort of strategic situations we are interested in, but there is an obvious way to turn this into a payoff structure that supports a signalling system: set a fixed, up-front cost for entry that satisfies condition [4.3], i.e. one that exceeds the benefit that a fly- by-nighter like Flaky would accrue before moving on. With enough time & experience for strategies to adapt to this costs-benefits regime, paying such an entry fee can plausibly evolve into a reliable signal of commitment.
This template arguably performs quite well with respect to the ideal explanatory virtues of signalling theory. It provides a clear link between commitment level and signalling strategy in equilibrium, since commitment directly determines sender payoffs. Because differential benefits are doing the strategic work, it predicts signals of significant fixed cost but arbitrary form: rituals, demands on time or resources, restrictions/requirements on behaviour, dress, and so-forth that can vary widely between communities. No psychological biases, exogenous constraints, punitive community norms, or other evolutionary pressures are required in principle. A community-erected ‘paywall’ is also scalable and evolvable assuming reasonable variation in commitment levels among potential senders: a small barrier to entry will immediately start filtering out the extremely flaky, incentivising the community to increase it up to some optimum level (beyond which profitable recruitment suffers). Coincidentally, for purposes of clarifying the literature, the general differential-benefit template is also similar to Ianacconne’s economic analysis of why strict churches are strong (Iannaccone 1992; 1994), allowing this view to be positioned as a differential cost-benefit fakeable signalling model. Other temporally extended assignments are possible which mix up the order of signal, cost, and benefit. For a final example consider an ‘investment signalling’ or ‘wise elder’ template:
spending time in one’s early life studying community-specific religious lore, so that authoritative signals of commitment to that community (and hence reliability as a cooperation partner) can be sent later. Although these later signals superficially look like they are hard-to- fake, the entire process can be seen as a long-run investment: pre-paying the costs for signalling commitment to local groups in a way that is far less expensive (at least in terms of opportunity cost) if you’re actually interested in belonging to just a small number – or one – of them. By being sensitive to these pre-paid signals, local communities can again discriminate between the genuinely (parochially) prosocial, and those who will to look for other options when the going gets tough.
6.4.4. Other possibilities
Of course, there even more modelling options available here. One possibility that was foreshadowed in chapter one, and which will be pursued further in chapters seven and eight, is that signalling strategies might have originated in less competitive strategic situations and environments. Under the right conditions, and because of constraints in trait specificity and/or evaluability, costless signalling à la the David Lewis signalling game might evolve for use in common-interest cases, but then ‘bleed over’ into environments more dominated by the prisoner’s dilemma-style conflicting interests. But this requires further discussion that will be left to later.
Vulnerability signalling should likewise be considered as a modelling option, though the operation of the model is different and would require more interpretation. In one sense, we can easily see something like a vulnerability ‘signal’ in submission rituals. Like a subordinate wolf baring its neck to the alpha male, a ritualistic submission to an authority (e.g. deep bowing or allowing them to place their foot upon one’s neck) is only costly if the receiver decides to reject the peace offering. But it is harder to recognise the rest of the vulnerability game payoff structure with respect to the other distributions of costs and benefits here. Also highly speculative is the projection of this game onto the public presentation of a religious authority (e.g. a holy man), whose schtick and promises are rewarded by community acceptance, but could be very costly if they were to turn on him as a fake. The problem with this interpretation is that it is hard to see what it would be for a holy man to not be a fake. The actual proportion of ‘high types’ would be 0, so unless we reconceptualise what is meant by ‘high type’ (e.g. by building in a subjective perception of benefit for the receivers, perhaps based on how good and falsely comforting a fake the holy man is) this will not work either.
This is not to say that vulnerability signalling cannot be linked to the cooperation problem or be placed in a religious context. Indeed, it would be surprising if it could not (as the payoff structure difference between it and intrinsic differential cost signalling are minimal). There may be possibilities for example with regards to temporally extended interpretations, perhaps modifying some of the cases considered above. However the point has largely already been made (that simple models can have many complex interpretations) and pursuing a more complex model in this way will have diminishing marginal returns for current purposes. 6.4.5. Summarising the options
Time to wrap this up. All the previous caveats about the difficulty of fitting models to real life cases should be kept in mind and updated when considering these more fleshed-out interpretations. For example, taking part in an intense ritual might generate commitment via our affective machinery, rather than displaying pre-existing commitment, and so something that seems to fit a Frankian hard-to-fake signal template (and produce broadly similar results) might be due to some other causal mechanism entirely. Likewise, these templates are by no means an exhaustive list, and variations on them (as well as entirely different signalling interpretations) are of course possible. With such caveats in mind, Table 6-1 summarises the seven that I have discussed.
Table 6-1: Seven application templates for models of religious signalling
Template Signalling model Context &
application Signal cost/honesty mechanism Special requirements Cognitive load (Sterelny) Hard-to-fake/index One-off/PAYG (dyadic interaction) Differential efficacy cost (cognitive effort) prohibiting false signal production
• Complex commitment signal conventions
• Relevant human cognitive limitations Emotional display (Frank) Hard-to-fake/index One-off/PAYG (ritual setting) Undifferentiated efficacy cost (pain, time, discomfort) prohibiting convincing false signal
• Affectively/physically demanding ritual conventions
• Relevant limitations in sustained affective ‘acting ability’ Subjective cost (Sosis) ‘Hybrid’ fakeable/handicap: subjective intrinsic differential cost One-off/PAYG (ritual setting) Undifferentiated signal costs made subjectively differential, directing behaviour as though strategic
• Perception bias correlated with both religiosity & prosociality
• Cooperation benefits must overwhelm actual signal cost in evolutionary dynamics Punishment Fakeable/handicap: intrinsic differential cost Extended/pay-later (initiation or maintenance) Undifferentiated signal with delayed differential strategic cost (punishment on defection) • Reliable punishment of defectors Bridge- burning Fakeable/handicap: intrinsic differential cost Extended/pay-later (persistent initiation markers)
Persistent signal with delayed differential strategic cost (3rd party rejection on defection) • Group-specific commitment conventions
• Rejection based on rival commitments
Investment Fakeable/handicap: intrinsic
differential cost
Extended/pre-pay (investment for later signalling)
Differential opportunity cost of learning investment (committed types only need invest once).
• Group-specific conventions of learnedness & respect
• Significant costs for learnedness investment Paywall Fakeable/handicap: intrinsic differential benefit Extended/earn-later (entry requirements for cooperative community) Undifferentiated signal/entry cost, but delayed differential accrual of cooperative benefits
• High signal/entry cost
The options here vary along several dimensions: signalling models, cost types and distributions, the temporal ordering of signal, cost, and benefit, and special requirements for a working RST interpretation. Each of these also vary in terms of their explanatory virtues, as discussed in the body text.
The recurring point to bear in mind is that they are not mutually exclusive, nor do they exclude other (non-signalling system) causal mechanisms. Many of them might be complimentary. So, if the targets of selection (the various tropes and contents of religious ritual and practice) are well-differentiated enough, then various signalling mechanisms might plausibly feature in explanatory causal complexes at the local community level (that we might even recognise as religions).
More generally, the scientific seriousness of RST will depend on how well its various moving parts are defined and combined: formal models and the mapping of them to real-world features e.g. actual costs and benefits (of the same currency), signal constraints, and relevant update/evolutionary mechanisms. The reader may baulk at the level of detail being recommended, but it is useful at the very least to acknowledge that such levels exist and can (in principle) become scientifically significant.