Morality constrains the default representation of what is possible

Significance

As humans, we think not only about what is, but also what could be. These representations of alternative possibilities support many important cognitive functions, such as predicting others’ future actions, assigning responsibility for past events, and making moral judgments. We perform many of these tasks quickly and effortlessly, which suggests access to an implicit, default assumption about what is possible. What are the default features of the possibilities that we consider? Remarkably, we find a default bias toward representing immoral or irrational actions as being impossible. Although this bias is diminished upon deliberative reflection, it is the default judgments that appear to support higher-level cognition.

Abstract

The capacity for representing and reasoning over sets of possibilities, or modal cognition, supports diverse kinds of high-level judgments: causal reasoning, moral judgment, language comprehension, and more. Prior research on modal cognition asks how humans explicitly and deliberatively reason about what is possible but has not investigated whether or how people have a default, implicit representation of which events are possible. We present three studies that characterize the role of implicit representations of possibility in cognition. Collectively, these studies differentiate explicit reasoning about possibilities from default implicit representations, demonstrate that human adults often default to treating immoral and irrational events as impossible, and provide a case study of high-level cognitive judgments relying on default implicit representations of possibility rather than explicit deliberation.

Human thought often involves alternative possibilities: how things could have been, might be, or may turn out. In other words, it involves modal cognition, a capacity to construct and reason over sets of nonactual events. Past research demonstrates a role for modal cognition in assessing causal responsibility (1–3), making moral judgments (4–6), inferring what others intend to communicate (7–9), and much more (10).

To date, research on modal cognition has not distinguished between implicit and explicit representations of possibility. In fact, past methods have overwhelmingly asked how we explicitly and reflectively represent and reason about different possibilities. We now know a great deal about how and when people engage in explicit counterfactual reasoning or reasoning about what else would have happened if a given alternative event had occurred (see ref. 11 for a recent review). There is also a small but growing body of research on how human adults deliberate and decide whether a given event is possible or impossible (12). In addition, there is an emerging literature on the neural substrates recruited when participants are instructed to engage in episodic counterfactual reasoning or simulation of possible future events (13, 14).

This research has been important for building an understanding of the way that human adults explicitly represent and reason about particular possibilities. But are these same processes recruited when humans make high-level judgments that are known to involve representations of possibilities? There is reason for doubt. Causal and moral judgments, for example, are often made quickly and effortlessly and appear early in human development (15, 16). Thus, these judgments are unlikely to rely exclusively on explicit deliberation about, or simulation of, alternative possibilities. Moreover, deliberative judgments about possibility often provide a poor fit to the patterns of high-level judgments they supposedly inform (see, e.g., ref. 17 on the dissociation between causal judgments and explicit counterfactual reasoning).

In light of this, an intriguing possibility is that people rely on a different kind of modal representation—one that is rendered quickly and automatically, and is based on a distinctive set of constraints. This could reflect a dissociable “system” or instead a graded contrast between default and deliberative judgments rendered by a common system. In either event, the aim of the present study is to test for a form of nondeliberative modal cognition, identify its signature properties, and determine its influence on higher-level cognition.

Critically, two independent research programs point toward a candidate signature property of the default representation: the relevance of prescriptive norms, such as morality and rationality. This stands in contrast to prior research on an adult’s deliberative reasoning about the (im)possibility of an event, which shows sensitivity to descriptive norms, such as the likelihood of the event occurring (e.g., ref. 12), but not to prescriptive norms (e.g., refs. 18 and 19). Thus, for instance, we tend to think “It isn’t possible for Lewis to open the safe” when it is highly unlikely he will guess the combination (a descriptive feature). In contrast, we tend not to think “It isn’t possible for Lewis to take the money” when it is sitting right in front of him, but belongs to somebody else (a prescriptive feature).

Remarkably, however, circumstantial evidence indicates that less explicit modal representations may be sensitive to both prescriptive and descriptive norms. That is, our default assessment of Lewis’s possible actions may exclude what would be wrong in the same way that it excludes what would be unlikely.

One source of evidence comes from studies of young children. Whereas children, much like adults, say that events that violate physical laws cannot happen (e.g., eating lightning), young children also say that events that violate moral or social norms cannot happen (e.g., stealing candy) (20–24). Not only do they judge these kinds of immoral events are “impossible,” but also they even say that such events would require “magic” (18, 19).

Additionally, some contemporary models of high-level cognition in adults argue that the underlying representation of possibility must be sensitive to both descriptive and prescriptive norms. This feature has played a critical role, for example, in understanding human causal reasoning (25, 26), modality in natural language (27), and judgments of force and freedom (28), among others (10). These models do not, however, provide evidence for how this modal representation is computed.

In sum, this evidence suggests that there may be a nondeliberative and early-emerging form of modal cognition that incorporates information about both descriptive and prescriptive norms. According to this hypothesis, the default representation would support not only judgments of what is “possible,” but also diverse modal judgments concerning what “ought,” “should,” “might,” “may,” or “could” be done. Moreover, the default representation would serve as input to diverse high-level cognitive judgments. We pursue these hypotheses in three sets of studies.

Study 1: Deliberative vs. Default Representations of Possibility

To investigate whether there are differences in more and less deliberative forms of modal cognition, we began by focusing specifically on judgments of possibility, asking participants to make judgments of whether various events were possible or impossible, while manipulating the amount of time available for this judgment. These judgments were made in the context of some background information, such as i:

• i)Josh is on the way to the airport to catch a flight for a hunting safari in Africa. He leaves with plenty of time to make it there, but his car breaks down on the highway. Now Josh is sitting in his car near a busy intersection and knows he needs to get to airport soon if he is going to catch his flight.

Participants read six different background contexts, and after each one they were shown a series of candidate events one at a time. As each event was presented, participants pressed a key to indicate whether they thought that the event was possible or impossible. Crucially, participants either were forced to make these judgments very quickly ($\le \mathrm{1,500}\hspace{0.17em}\text{ms}$) or were asked to reflectively deliberate on the possibility of each event ($\ge \mathrm{1,500}\hspace{0.17em}\text{ms}$). We used a total of 144 different events, which were designed to fall into five categories: 48 ordinary events that did not violate any norms, e.g., ii, a; 24 events that violated statistical norms, e.g., ii, b; 24 events that violated physical laws, e.g., ii, c; 24 events that violated moral rules, e.g., ii, d; and 24 events that violated norms of rationality, e.g., ii, e. (The a priori categorization of these events was confirmed in independent ratings of the morality, probability, and rationality of each of the events; SI Text, Event Ratings and Fig. S1.)

• ii)Is it possible or impossible for Josh to [(a)] hail a taxi at the intersection, [(b)] fix his car by banging on it, [(c)] teleport himself to the airport, [(d)] sneak onto public transportation, or [(e)] sell his car for a ride to the airport?

Fig. S1.

Participants’ ratings of the probability, rationality, and morality of the 144 events, split by a priori event category. Error bars depict $\pm$1 SEM.

This paradigm uses judgments of whether an event is possible as an indicator of whether participants represent that event as being available in the context. By changing the amount of time participants have to deliberate, we ask whether (and how) participants’ representation of the set of available events changes as a result of more or less deliberation.

We first used a series of linear mixed-effects models to test whether participants’ judgments of the possibility of the five different kinds of events differed across time pressure/delay conditions (SI Text, Statistical Approach). We observed an effect of whether participants were reflectively deliberating, ${\chi }^2(1)=20.653$, $P<0.001$, and an effect of event-type ${\chi }^2(4)=860.48$, $P<0.001$. Critically, however, these main effects were qualified by a deliberation $\times$ event-type interaction, ${\chi }^2(4)=64.093$, $P<0.001$. We decomposed this interaction, using a series of generalized linear models. These revealed that deliberation did not significantly affect participants’ judgments of the possibility of ordinary events ($z=-0.164$, $P=0.87$) or improbable events ($z=-1.056$, $P=0.291$). Although not predicted, participants more tended to judge that events that violated physics were impossible after deliberating ($M=85.22$, $\text{SD}=17.44$), than when forced to answer quickly ($M=79.37$, $\text{SD}=15.82$), ($z=-6.628$, $P<0.001$). Critically, however, we found that participants’ judgments of the possibility of immoral and irrational events were affected by deliberation in the opposite direction. They tended to judge that immoral events were impossible when they were not able to reflectively deliberate ($M=37.30$, $\text{SD}=30.21$), but more judged them to be possible after deliberating ($M=21.41$, $\text{SD}=31.26$), ($z=5.423$, $P<0.001$). Similarly, they tended to judge that irrational events were impossible more when they were not able to deliberate ($M=42.10$, $\text{SD}=25.89$), than after deliberating ($M=37.86$, $\text{SD}=30.89$), ($z=1.963$, $P<0.05$) (Fig. 1).

Fig. 1.

Judgments of impossibility for five different kinds of events when participants made judgments after deliberating (solid bars) or without time to deliberate before responding (shaded bars). Error bars indicate $\pm$1 SEM.

In short, we find evidence that default representations of what is possible differ from more reflective representations in that prescriptive norms selectively constrain the default representation of possibility. Moreover, two key contrasts suggest an especially strong role for moral norms: Judgments of the possibility of immoral events were more affected by deliberation than those of either improbable events ${\chi }^2(1)=28.035$, $P<0.001$, or irrational events ${\chi }^2(1)=24.937$, $P<0.001$, despite the fact that all of these events involved clear norm violations.

Event Ratings

Participants were randomly assigned to rate the probability, rationality, or morality of each of the 144 events. Those who rated the probability of each action answered the question “How likely is [Agent] to [event]” on a scale from 1 (“very likely”) to 5 (“very unlikely”). Those who rated the rationality of each action answered the question “How irrational would it be for [Agent] to [event]” on a scale from 1 (“totally rational”) to 5 (“totally irrational”). Those who rated the morality of each action answered the question “How morally wrong would it be for [Agent] to [event]” on a scale from 1 (“totally morally fine”) to 5 (“totally morally wrong”). In all cases, participants had an option to respond that the question was “not applicable.” Participants saw all six of the contexts in random order and, after reading each context, provided a rating of each of the 24 events for that scenario (8 ordinary, 4 immoral, 4 improbable, 4 irrational, and 4 impossible) in random order.

All trials on which participants provided “NA” responses were excluded, and the remaining responses were analyzed to ensure that the a priori categorization of the events was confirmed (Fig. S1). These ratings confirmed our categorization. As a group, “ordinary” events were judged to be less unlikely than improbable events, $t(44)=-13.789$, $P<0.001$, $d=4.07$; less irrational than irrational events, $t(38)=-16.536$, $P<0.001$, $d=5.23$; and less morally wrong than immoral events, $t(34)=-20.375$, $P<0.001$, $d=6.79$. Improbable events, in addition to being perceived as more unlikely than ordinary events, were also less irrational than irrational events, $t(38)=-6.810$, $P<0.001$, $d=2.15$, and less morally wrong than immoral events, $t(34)=-17.893$, $P<0.001$, $d=5.99$. Unsurprisingly, impossible events were judged to be more improbable than any other group of events, including improbable events, $t(44)=3.888$, $P<0.001$, $d=1.15$, and more irrational than any other group of events, including irrational events, $t(26.27)=5.163$, $P<0.001$, $d=1.63$. They were not judged to be particularly immoral and much less immoral than immoral events, $t(34)=-10.143$, $P<0.001$, $d=3.38$. Immoral events, in addition to being more immoral than all other events, were judged to be relatively unlikely and irrational. Critically, however, they were not judged to be any more unlikely than improbable events, $t(44)=0.028$, $P=0.978$, $d<0.01$, and were judged to be marginally less irrational than irrational events, $t(38)=-1.865$, $P=0.070$, $d<0.59$. Similarly, irrational events, although unsurprisingly rated as unlikely, were not judged to be more unlikely than improbable events, $t(44)=1.637$, $P=0.110$, $d<0.483$. They were, however, rated as less morally wrong than immoral events, $t(38)=-10.488$, $P<0.001$, $d<3.50$. Accordingly, the patterns observed for the immoral and irrational events, but not for the improbable or impossible events, must be explained by their elevated levels of immorality and irrationality, rather than the extent to which these events violated descriptive norms.

Statistical Approach

When possible, the primary analyses were conducted with generalized linear mixed-effects models (37) in R (38). The significance of each effect was determined by comparing a model that included the relevant term in the model (as well as other factors that were not currently being investigated) to a model that did not include that term (but did include the other factors not under investigation). The effect was determined to be significant if the fit of the model including the relevant term differed significantly from the fit of the model that did not include that term (39).

Study 2: Does This Default Generalize Across Modal Judgments?

Linguistically, different modal auxiliaries (e.g., might vs. ought) are known to select different sets of possibilities (7, 8). All of these modal auxiliaries are alike, however, in that they select some set of contextually relevant possibilities. This is also true in nonlinguistic cognition: We think differently about what a person could or might do than we do about what they should or ought to do. In all of these cases though, these judgments require us to reason over the set of events that are represented as being available in the context. Accordingly, we can ask whether distinct modal judgments are supported by distinct default sets of possibilities or whether instead they share a common set of default possibilities and become differentiated only after additional processing.

We propose that there is a common mechanism responsible for rapidly constructing a default representation of what is possible, which supports diverse cognitive functions. This proposal predicts that any deviation from the default representation will require additional processing, and thus the selection of different sets of possibilities for different modal judgments is likely the result of more deliberative cognition. This prediction is testable: We can ask whether the correlation between judgments of what a person could do, might do, ought to do, etc., is higher when people do not have time to reflect (and thus rely more on a default representation of possibility) and lower when people do have time to reflect before making a modal judgment.

An obvious alternative prediction is that reducing the amount of time that participants have to respond should increase the noise observed in their responses. This predicts that modal judgments will instead be less correlated with one another when they have to be made extremely quickly (compared with when they are made with time to reflect). We hypothesized that this effect would be outweighed by the influence of a common default modal representation.

Motivated by the results of study 1, we further hypothesized that the default representation underlying diverse modal judgments would be sensitive to prescriptive norm violations. In other words, we hypothesized that under time pressure participants’ judgments of what a person could do, may do, might do, etc., would be sensitive to whether that action is immoral or irrational. This common sensitivity to prescriptive norms should thus be largely responsible for heightened similarity among the default representations elicited by diverse modals.

Following the basic procedure introduced in study 1a, we collected deliberative vs. nondeliberative judgments of what agents may do (study 2a), might do (study 2b), could do (study 2c), ought to do (study 2d), or should do (study 2e). We also included judgments of what it is possible for them to do (from study 1). These different modal terms are known to express different kinds of modality (metaphysical, circumstantial, deontic), and the overall patterns of responses demonstrate that participants tracked these differences (SI Text, Differences Across Modal Auxiliaries and Fig. S2).

Fig. S2.

Mean judgments for the six different modal terms for events where there were no salient norm violations (gray), events where only descriptive norm violations were salient (orange), and events where prescriptive norm violations were relevant (tan). Error bars depict $\pm$1 SEM.

Our analyses treat the 144 different events as the primary unit of analysis. For each of the six modal judgments, we calculated the proportion of the time that the modal judgment was accepted vs. rejected for each event, doing so independently for trials when participants had time to reflect and when they did not. These scores reflect the likelihood that an event will be included in the set of possibilities relevant to each type of modal concept (e.g., what should be done, what could be done, and so on). We then calculated the correlation between the events’ representations for each possible pair of different modal judgments, both when participants were given time to reflect before answering and when they were not.

As hypothesized, modal judgments were significantly more correlated when participants did not have time to reflect ($M_r=0.893$, ${\text{SD}}_r=0.060$) than when they did ($M_r=0.830$, ${\text{SD}}_r=0.101$), $t(14)=5.131$, $P<0.001$, $d=0.754$ (Fig. 2). This finding supports the view that participants relied on a common default representation of the set of available events when answering quickly. By contrast, when participants reflected on each judgment, their responses were more strongly dictated by unique features of each modal concept and thus became less correlated with one another.

Fig. 2.

Correlation coefficient of each pair of modal questions, either when participants were forced to answer quickly (left) or when they were given time to reflect before answering (right).

We next asked whether a principal dimension on which default judgments became more similar was sensitivity to prescriptive norms. To do this, we divided the events that participants judged into two categories: those for which descriptive norms were primarily relevant (the improbable and physically impossible events) and those for which prescriptive norms were primarily relevant (the irrational and immoral events). We then again calculated the correlation between each pair of modal judgments, first focusing on events where descriptive norms were relevant and then, second, focusing on events where prescriptive norms were relevant. This analysis allows us to ask whether it was the presence of one of these kinds of norms that was responsible for the difference in correlations when participants reflected before answering.

Analyzing these correlations with linear mixed-effects models, we found an interaction effect between the type of norms that were relevant and whether or not participants had time to reflect, ${\chi }^2(1)=24.323$, $P<0.001$. Decomposing this interaction, we first focused on modal judgments made in scenarios for which descriptive norms were primarily relevant (e.g., winning the lottery). Here, we observed a modest decrease in the correlation between the modal judgments when participants reflected before answering ($M_r=0.873$, ${\text{SD}}_r=0.055$) from when they answered before reflecting ($M_r=0.908$, ${\text{SD}}_r=0.042$), $t(14)=5.336$, $P<0.001$, $d=0.724$. Next, we focused on judgments made in scenarios for which prescriptive norms were relevant (e.g., theft) and found a much larger decrease in the correlations between the modal judgments made when participants reflected before answering ($M_r=0.278$, ${\text{SD}}_r=0.265$), compared with when they answered before reflecting ($M_r=0.628$, ${\text{SD}}_r=0.145$), $t(14)=8.841$, $P<0.001$, $d=1.642$ (Fig. 3). A corollary finding is that deontic modals show less of a difference between deliberative and reflective responses than metaphysical or circumstantial modals (SI Text, Within-modal correlations).

Fig. 3.

Correlation coefficient for each pair of modal judgments either for items where descriptive norms were primarily relevant (Left) or for items where prescriptive norms were primarily relevant (Right). Correlations between modal judgments made without time to reflect are plotted on the left-hand side of each panel; correlations between modal judgments made after reflecting are plotted on the right-hand side of each panel.

Together, these analyses indicate that prescriptive norms constitute a principal dimension differentiating specific modal concepts from their common default. In contrast, the use of descriptive norms appears to be relatively more consistent across distinct modal concepts, even upon reflection. This pattern is most evident for moral norms. Focusing specifically on the immoral events, which clearly distinguish different kinds of modal reasoning, we created a correlation matrix of all of the different judgments. This approach revealed that all of the different modal judgments were highly correlated when participants were not able to reflect ($0.558\le r\le 0.919$), but were not when participants took time to reflect ($-0.016\le r\le 0.572$) (Fig. 4). A corollary finding is that participants’ nonreflective modal judgments were all highly correlated with reflective judgments of what agents ought to do ($r_{Mean}=0.465$), but not other reflective judgments ($-0.091<r_{Mean}>0.264$). Put simply, when answering quickly, modal judgments become more ought-like (box in Fig. 4).

Fig. 4.

Graphical depiction of the correlation matrix for speeded and reflective modal judgments of events that involved immoral actions. Squares within the black box (bottom left) depict correlations of speeded modal judgments with reflective judgments of ought.

In summary, we have demonstrated thus far that default modal judgments show a common sensitivity to both descriptive and prescriptive norms. In contrast, reflective modal judgments can deviate away from this default, allowing for some kinds of modal cognition (e.g., thoughts of what could happen) to focus primarily on what is physically possible, without regard to immorality or irrationality.

Differences Across Modal Auxiliaries

It is widely known that some modal terms demonstrate a preference for more “metaphysical” readings (e.g., possible, could) and thus primarily concern the possible ways things could occur. Other modal terms demonstrate a preference for a “circumstantial” reading (e.g., might, may) and thus concern the way that things would normally proceed in the circumstances. Others instead demonstrate a preference for “deontic” readings (e.g., ought, should) and concern what would be morally good or rationally appropriate. The differences between these different modal terms was reflected in participants’ judgments. Collapsing across speeded and reflective responses, judgments of what was possible and what could occur were most concerned with descriptive norms but not prescriptive norms ($t$s $>3.62$, $P$s $<0.001$, $d$s $>0.738$). Judgments of might and may were roughly equally sensitive to both prescriptive and descriptive norms ($t$s $<1.95$, $P$s $>0.05$, $d$s $<0.397$). And judgments of should and ought were specifically sensitive to prescriptive norm violations but less sensitive to descriptive norm violations ($t$s $<-3.15$, $P$s $<0.003$, $d$s $>0.644$), as would be expected (Fig. S2).

Within-Modal Correlations

Given these differences in the modal terms, a natural corollary of our findings in study 2 is that for events involving prescriptive norm violations, deontic modal judgments made under time pressure and after deliberation should be highly correlated, because moral norms will influence participants’ judgments either way. Thus, for example, reflective judgments of whether an immoral or irrational action ought to be done should be highly correlated with speeded judgments of whether that action ought to be done. By contrast, speeded and reflective judgments of circumstantial modals such as might should be less correlated, and more metaphysical modals such as possible should be even less correlated. Consistent with this prediction, we found that the similarity of speeded and reflective judgments of deontic modals were highly correlated ($r_{ought}=0.537$, $r_{should}=0.531$), that circumstantial modal judgments were relatively less correlated ($r_{may}=0.407$, $r_{might}=0.390$), and that metaphysical modal judgments were even less correlated ($r_{could}=0.358$, $r_{possible}=0.376$).

Study 3: Default Modal Representations in High-Level Cognition

Existing evidence suggests that the modal representations used in other domains of high-level cognition—for instance causal attribution, language comprehension, and moral judgment—are similar to the default representations identified in studies 1 and 2. For instance, causal judgments of an event are sensitive to the moral status of alternative events (e.g., refs. 25 and 29), which is consistent with the demonstrated role of prescriptive norms in default modal representations. This current evidence is indirect, however, and the role of alternative events in causal cognition remains controversial (e.g., ref. 30). Accordingly, study 3 provides a direct test of the hypothesis that the default representations of the set of events available in a given context are exported to other cognitive domains.

To do this, we used judgments of “force” as a case study. This is an ideal test because empirical (28, 31) and theoretical (32, 33) studies concur that judgments of force rely on representations of alternative possibilities. Put simply, if an agent was “forced” to act in some way, this implies that there was no relevant alternative; and if he or she was not forced, this implies that some alternative existed. We therefore asked, when people decide whether an agent was forced to do something, do they rely on a set of possibilities more like a deliberative set or instead on the implicit set that we identified in study 1?

Participants in this study once again read the six different background contexts used in the previous studies. However, in this case, the agent asked another person for advice about what to do. This person then directly suggested that the agent pursue one of the ordinary, immoral, irrational, or improbable events used previously (ii, a–e), as in iii:

• iii)Josh calls his father who lives a few states away and tells him about his problem. Not really knowing how to help him, Josh’s father makes a suggestion. His father says Josh could sneak onto public transportation.

Participants were then told that, notwithstanding this advice, the agent decided to pursue some specified alternative course of action, as in iv:

• ii)Josh ignores his father’s suggestion and decides to book the next available flight, even though it is quite expensive.

Within each context, the agent was always described as pursuing the same course of action, regardless of the advice given.

After reading all of this information, participants were asked whether they agreed or disagreed that the agent was forced to do the action he or she actually pursued, as in v:

• i)Josh was forced to book the next available flight.

Following previous work on judgments of force (28, 32, 33), participants should judge the agent to have been forced only if they believe that there were no other options available to the agent—in other words, if the set of relevant alternative possibilities was empty. Thus, in the context of the specific advice given to the agent, participants should agree that the agent was forced only if they did not represent the proposed option as an available possibility.

Crucially, in all cases, the judgments of force were made with unlimited time available. Thus, our question is whether deliberative judgments of force for each scenario are better predicted by deliberative or speeded judgments of possibility for the same scenario in study 1. We hypothesized that force judgments would be best predicted by speeded possibility judgments, reflecting a dominant influence of the default representation of the set of available events.

To analyze these data, we computed the average agreement that the agent was forced to do a given action when the alternative proposed was one of the events used in the previous studies. We then first asked whether participants’ nonreflective judgments were predictive of judgments of force in a simple linear model and found that they were, $F(\mathrm{1,118})=46.12$, $P<0.001$, ${\eta }^2=0.281$ (Fig. 5). Next, we asked whether reflective or nonreflective judgments of possibility were a better predictor of judgments of force by comparing a series of linear mixed-effects models. We found that including reflective judgments of possibility did not significantly improve a model that already included nonreflective judgments of possibility, ${\chi }^2(1)=0.030$, $P=0.863$. In contrast, including nonreflective judgments of possibility did significantly improve a model that already included reflective judgments of possibility, ${\chi }^2(1)=27.640$, $P<0.001$. Thus, above and beyond reflective judgments of what is possible, the default representation of possibility is predictive of deliberative judgments of force.

Fig. 5.

Relationship between the implicit representation of possibility and high-level judgments of whether or not an agent was forced to do a given action in six different contexts. Shaded areas represent 95% confidence intervals.

We next asked whether morality had an effect on judgments of force [as found in previous work (28, 31)] and whether this effect was mediated by nonreflective judgments of possibility. We found that this was the case: Participants judged agents to be more forced when the proposed alternative was immoral vs. ordinary, $t(70)=7.309$, $P<0.001$, $d=1.827$. Moreover, this effect was mediated by nonreflective judgments of possibility, 95%CI [0.018, 0.542], $P=0.04$. A similar pattern of results was also observed for events that involved violations of rational norms ($t(70)=5.462$, $P<0.001$, $d=1.366$; 95%CI [0.045, 0.864], $P=0.04$). These results fit well with ongoing work that offers a semantics for force according to which prescriptive norms impact the set of actions available to an agent (34). (See SI Text, Study 3b: Can Inferences About Agents’ Desires Explain the Impact of Norms on Judgments of Force? for a test of an alternative semantics for force that proposes to account for the impact of norms through inferences about the agents’ desires.)

Study 3b: Can Inferences About Agents’ Desires Explain the Impact of Norms on Judgments of Force?

The results of study 3a provide evidence that the impact of norms on deliberative judgments of force are best explained by implicit, default representations of possibility. This account relies on the relatively standard assumption that alternative possibilities play a central role in judgments of force, e.g., refs. 33 and 34; see ref. 35 for a formal exposition of the semantics of force along these lines.

An alternative possibility, however, is that prescriptive and descriptive norms instead influenced judgments of force by changing participants’ inferences about the agent’s desires (and thus the default representation of possibilities did not play a direct role). According to this explanation, when an immoral or statistically unlikely event was proposed to the agent, participants were likely to infer that the agent had a strong desire to not do that action, which in turn led them to see the agent as forced to do some other action instead. (We thank an anonymous reviewer for raising this alternative way of accounting for the impact of norms on judgments of force.) If correct, this alternative explanation would undermine the earlier evidence that the default representations of possibility play a role in high-level deliberative judgments of force.

To test this alternative proposal, we collected data on participants’ inferences about the agents’ desires to not do the proposed action. Similar to study 3a, participants first read the six different background contexts used in the previous studies and then read a continuation of that context in which another person proposes that the agent pursue one of the ordinary, immoral, irrational, or improbable events. Participants were then told that the agent always decided to pursue some specified alternative course of action, regardless of the advice given. After reading each background context, participants rated their agreement on a scale from a 1 (“completely agree”) to 5 (“completely disagree”) with a statement that the agent did not want to do the alternative proposed action. For example, in the scenario in which Josh’s father suggests that he could sneak onto public transportation, but Josh ignores his suggestion and books the next available flight, participants were asked whether they agreed or disagreed with the following statement: Josh did not want to sneak onto public transportation.

According to this alternative explanation, participants should perceive the agent as not wanting to do the actions that involved violations of prescriptive or descriptive norms and, moreover, that this difference in inferred desire should explain the pattern of force judgments from study 3a. We investigated whether this was the case by computing the average desire ratings for each of the proposed events and then asking how these averages related to (i) the different categories of events and (ii) previous judgments that the agent was forced to do the alternative. First, we found that participants’ judgments that the agent did not want to do the action were significantly influenced by the kind of action that was proposed, $F(\mathrm{3,116})=33.73$, $P<0.001$, ${\eta }^2=0.466$. Specifically, participants tended to judge that the agent wanted to do proposed actions less when they were immoral ($M=1.34$, $\text{SD}=0.17$) or irrational ($M=1.55$, $\text{SD}=0.26$) than when the they were ordinary ($M=1.89$, $\text{SD}=0.47$), $t$s $<-3.88$, $P$s $<0.001$, $d$s $>0.81$. However, we also found that participants judged that the agent did want to do the improbable actions ($M=2.48$, $\text{SD}=0.59$). Presumably, this pattern occurred because these actions had highly good outcomes but were out of the agent’s control, e.g., fixing a car by banging on it.

More importantly, we next asked whether these judgments of the agents’ desires were a better predictor of judgments of force than the implicit, default representation of possibility. To do this, we first built a linear mixed-effects model that included the default representation of the possibility of each event and then asked whether this model was improved by adding in participants’ judgments of whether the agent desired to not do the proposed action. This was not the case, ${\chi }^2(1)=0.760$, $P=0.383$, suggesting that inferences of the agent’s desire added little predictive value above and beyond that already accounted for by implicit representations of possibility. Conversely, however, implicit representations of possibility did improve a model that included only judgments of what the agent desired to not do, ${\chi }^2(1)=52.92$, $P<0.001$.

Finally, we asked whether assessments of agents’ desires could at least account for the specific impact of prescriptive norms on judgments of force. In answer to this question, we found that perceptions of the agents’ desires did not mediate the impact of morality on judgments of force (95% CI [−0.154, 0.081], $P=0.57$). Similarly we also found that perceptions of the agents’ desires did not mediate the impact of rationality on judgments of force (95% CI [−0.098, 0.058], $P=0.53$). In short, we did not find evidence that perceptions of the agents’ desires could explain participants’ judgments that the agent was forced to do a given action when the proposed alternative was immoral or irrational. These analyses provide further evidence for the critical importance of default representations of possibility in high-level deliberative judgments.

Discussion

We find evidence for a default representation of what is possible, i.e., a set of events, specific to a context, that constitutes a starting point for understanding nonactual alternatives. A hallmark of this default representation is that it tends to exclude immoral actions. First, we found that time pressure makes people less likely to judge it possible to act immorally or, to a lesser extent, irrationally. Second, we found that this effect is reflected across diverse modal concepts: what a person ought to do, may do, could do, should do, or might do. These judgments appear to share a common default basis and become more differentiated only after deliberative processing. Finally, as a case study of the role of implicit modal cognition in reflective high-level judgments, we found that default modal judgments are better predictors than reflective modal judgments of people’s decisions about whether a person was forced to act in a particular way. A mediation analysis indicates that, to the extent that we are forced to act morally, it is because our default representation of what is possible tends to exclude immoral actions.

Although all forms of modal cognition involve constructing and reasoning over sets of possibilities, it is remarkable that they apparently begin with a common default template. By analogy, all forms of social cognition—whether about in-groups or out-groups, cooperators or competitors—involve reasoning about sets of other people, but it would be remarkable to discover that under time pressure, all forms of social cognition begin with a representation of the very same individuals. In modal cognition, much like in social cognition, one might have initially thought that even rapid categorizations of what is possible for a person to do and what a person ought to do would depend on separate processes that pick out separate events. After all, most people would say that it is possible to run a red light at an empty intersection, but not that one ought to. In contrast to this intuition, our findings suggest that it takes time to realize that it is possible to run a red light. We begin with a default representation of possibility that tends to exclude this action, along with other immoral or irrational acts.

The purpose of this default mechanism is a key area for further study. One appealing possibility is that the default representation serves the function of proposing actual candidate actions during decision making—in other words, it helps us to construct a choice set (35). It is natural to suppose that screening out immoral or irrational actions from one’s own decision making would tend to be adaptive.

Importantly, the default representation we have uncovered bears a striking resemblance to the pattern of possibility judgments observed early in human development (18, 19, 21, 24). In both cases, judgments of possibility are sensitive not only to descriptive norms but also to prescriptive ones. The similarity of these patterns presents the intriguing possibility that the modal representation observed early in development is retained in adult cognition alongside a separate later-developing capacity for deliberatively reasoning about possibilities. Future work should continue to explore this connection directly and ask which other factors serve to constrain both of these representations of possibilities.

Finally, it is worth noting a potential connection between our research and the widespread, surprising presence of moral norms in disparate corners of the human mind: e.g., in causal reasoning and mental state attribution (10, 25, 26, 29, 36). These processes also involve reasoning over a set of possibilities. Given the present results, an exciting possibility is that morality shapes how we think about many things because it constrains the very possibilities that come to mind.

Materials and Methods

Participants.

All participants were recruited from Amazon Mechanical Turk (www.mturk.com) through TurkPrime (www.turkprime.com), which was used to prevent repeat participation across the studies. Sample sizes and demographic information are as follows: study 1a (judgments of possibility), 498 participants ($M_{age}=33.85$, ${\text{SD}}_{age}=10.82$, 233 females); study 1b (event ratings), 61 participants ($M_{age}=31.53$, ${\text{SD}}_{age}=7.87$, 25 females); study 2a (ought), 301 participants ($M_{age}=33.40$, ${\text{SD}}_{age}=9.87$, 147 females); study 2b (might), 201 participants ($M_{age}=33.79$, ${\text{SD}}_{age}=10.20$, 92 females); study 2c (could), 304 participants ($M_{age}=34.87$, ${\text{SD}}_{age}=11.32$, 144 females); study 2d (may), 200 participants ($M_{age}=32.80$, ${\text{SD}}_{age}=9.21$, 86 females); study 2e (should), 297 participants ($M_{age}=34.05$, ${\text{SD}}_{age}=9.86$, 153 females); study 3 (force judgments), 400 participants ($M_{age}=38.05$, ${\text{SD}}_{age}=12.73$, 218 females); and study 3b (desire judgments), 400 participants ($M_{age}=36.61$, ${\text{SD}}_{age}=11.80$, 200 females). These studies were approved by the Harvard University Institutional Review Board, IRB14-2016, and informed consent was acquired from all participants.

Access to Data, Analysis Code, and Materials.

All materials used to conduct these studies and analyze the results are available at github.com/phillipsjs/implicitModality. Studies 1 a and b and 2 a–e were conducted through Testable (www.testable.org) and study 3 a and b were conducted in Qualtrics. Stable links to experiments are as follows: study 1 a and b, (possibility) testable.org/t/3111b018 and (event ratings) testable.org/t/31041bbc; study 2 a–e, (ought) testable.org/t/31092d3e, (might) testable.org/t/3120b96c, (could) testable.org/t/31bbed30, (may) testable.org/t/31f78b37, and (should) testable.org/t/317a2d7c; and study 3 a and b, (force) goo.gl/6POVuY and (desire) https://goo.gl/30OjG1.

Exclusion Criteria and Supplemental Analyses

Studies 1b and 2 a–e.

Trials on which participants did not respond were excluded from the analyses. Subsequently, each participant’s average response time (excluding outlier responses defined as $>\mathrm{6,000}\hspace{0.17em}\text{ms}$) was computed. All data from a participant were dropped if a participant’s average response time was lower than 800 ms in the speeded condition or lower than 1,000 ms in the reflective condition ($\approx 6\%$ of participants in study 1a and $\approx 8\%$ in study 2 a–e). Additionally, data from all trials on which a response was given in less than 500 ms were dropped ($\approx 1\%$ in study 1a and study 2 a–e), as were data from reflective trials on which the response was given in less than 1,500 ms ($\approx 27\%$ in study 1 and study 2 a–e).

Supplemental analyses.

To ensure that the key differences we observed did not arise from a selection effect produced by the unequal exclusion of data from the reflective condition, we reanalyzed the data without the separate exclusion criteria for the reflective condition. Whereas relaxing these exclusion criteria obviously reduces the difference between the two conditions, we still continued to observe the key results. In study 1a, we again observed an effect of whether participants were asked to reflectively deliberate, ${\chi }^2(1)=18.403$, $P<0.001$; an effect of event type, ${\chi }^2(4)=880.23$, $P<0.001$; and critically, a deliberation $\times$ event-type interaction, ${\chi }^2(4)=76.02$, $P<0.001$. This interaction effect continued to be driven in part by the fact that participants tended more to judge immoral events were impossible when they responded quickly ($M=37.30$, $\text{SD}=30.21$) than when they responded after deliberating ($M=23.23$, $\text{SD}=31.10$) ($z=4.942$, $P<0.001$). Similarly, they tended to judge that irrational events were impossible more when answering quickly ($M=42.10$, $\text{SD}=25.89$) than after reflecting ($M=37.33$, $\text{SD}=28.54$) ($z=2.115$, $P<0.05$).

Moreover, in study 2, we continued to find that modal judgments were significantly more correlated when participants did not have time to reflect ($M_r=0.892$, ${\text{SD}}_r=0.059$) than when they did ($M_r=0.858$, ${\text{SD}}_r=0.106$), $t(14)=2.341$, $P=0.035$, $d=0.403$. We additionally continued to observe an interaction effect between the type of norms that were relevant and whether or not participants had time to reflect, ${\chi }^2(1)=8.545$, $P=0.003$. Decomposing this interaction, we found that for descriptive-norm events, there was no decrease in the correlation between the modal judgments when participants reflected before answering ($M_r=0.912$, ${\text{SD}}_r=0.052$) from when they answered before reflecting ($M_r=0.908$, ${\text{SD}}_r=0.042$), $t(14)=0.491$, $P=0.631$, $d=0.077$. Next, we focused on judgments made when prescriptive norms were relevant and found a much larger decrease in the correlations between the modal judgments made when participants reflected before answering ($M_r=0.438$, ${\text{SD}}_r=0.290$), compared with when they answered before reflecting ($M_r=0.628$, ${\text{SD}}_r=0.145$), $t(14)=3.569$, $P=0.003$, $d=0.831$.

Finally, in study 3, we continued to find that deliberative judgments of force were better predicted by speeded judgments of possibility. Specifically, including reflective judgments of possibility did not significantly improve a model that already included nonreflective judgments of possibility, ${\chi }^2(1)=0.097$, $P=0.756$. In contrast, including nonreflective judgments of possibility did significantly improve a model that already included reflective judgments of possibility, ${\chi }^2(1)=20.583$, $P<0.001$.

Studies 1a and 3 a and b.

No data were excluded from these experiments as there was no time manipulation, and thus no supplemental analyses were conducted.