Moving from theory to evidence, there has been no empirical research investigating whether adversarial dialogue is superior. However, there are studies involving manipulation of conflict which show increased liking both of stories with high “emotional conflict” (Diener & Woody, 1981), high stakes goals (Brewer & Lichtenstein, 1981), and stories with significant obstacles to goal achievement (Jose, 1988). In these studies, structural and dramatic conflict are not overtly differentiated. However, the studies of Brewer and Lichtenstein, and Jose give enough detail to suggest the former manipulates structural conflict, while the latter manipulates both structural and dramatic conflict.

More research examines the mechanisms of positive audience response to stories containing unmanipulated conflict.

Conflict is used in story research to create suspense (Comisky & Bryant, 1982; Doust, 2015; Gerrig & Bernardo, 1994; Madrigal et al., 2011; Vorderer & Knobloch, 2000; Zillmann et al., 1975), and research show that suspense leads to story enjoyment (Brewer & Lichtenstein, 1981; Madrigal et al., 2011), however, we lack experimental evidence showing the whole pathway: increasing conflict increases suspense, in turn increasing story enjoyment. Similarly, though research on the ADT of drama has shown that the resolution of a story conflict in favour of a liked character leads to enjoyment (Grizzard et al., 2023; Hoffner & Cantor, 1991; Raney & Bryant, 2002; Zillmann, 2000; Zillman & Cantor, 1977), and appreciation research has shown that a mixed valence story resolution for a liked character leads to appreciation (Bennie et al., 2020; Grizzard et al., 2021; Knop-Huelss et al., 2020; Lewis et al., 2014), none of this research involved manipulation of conflict.

So, in this literature there are a number of gaps, including experimental research manipulating isolated structural, and dramatic conflict. The latter variable is theoretically significant, because dramatic conflict is central to the dramatic theory, and part of the causal pathway of the traditional theory of conflict. As a moderating variable of unknown strength, dramatic conflict’s importance is less certain in the psychological theory.

In addition to these theoretical issues, the experiments presented here also have a practical purpose: To investigate how to improve dialogue — an important component of a well written story (Axelrod, 2013; Bell, 2014; Frey, 2010; Hough, 2015; McKee, 2016; Stein, 1995).

Table 6 gives descriptive statistics from Experiment 1.

Table 7 gives inferential statistics from Experiment 1, showing the manipulation of conflict led to a highly significant change in perceived conflict, with a large to very large effect size, whereas it had no significant effect (at a p < .05 level) on the measure of quality (total PSQI) or Entertainment, with small to very small effect sizes.

The loss of data, through losing Condition 7 (see Table 3) decreased the experiment’s power, and introduced potential bias by creating a correlation between conflict level and genre/story type — the low conflict stories were less likely to be romance stories, and the high conflict stories less likely to be fantasy stories. To address this, we created a multilevel model, with random intercepts for the three genres/story types, so controlling for the effect of genre/story type (see ESM 1 at Berks, 2026). The idea was both to reduce the noise in the PSQI ratings and thus increase power to detect an effect of conflict and get a sense of how much genre/story type might be affecting the results. For both the total PSQI and the Entertaining item, the intercept only model showed an intraclass correlation coefficient for the grouping variable (individuals) of .00, suggesting, somewhat surprisingly, that individual differences made no difference to either rating. The fixed effects for both the total PSQI and the Entertaining item suggested any effect of genre/story was insignificant.

While this study concerns the effect of conflict in stories in general, the effect of this conflict may vary across genre/story type, and so post hoc analyses of these effects are included in ESM 1 (Berks, 2026), showing no variation across genre.

In Experiment 1, we increased the conflict in the dialogue by following recommendations in the literature (Patterson, 2013; Ware et al., 2012), which increased the perceived level of conflict between the characters. This increase in perceived level of conflict was both statistically significant and — using the criteria of a Cohen’s d of more than 0.8 (Cohen, 1988) — had a large to very large effect size. Yet participants did not rate the stories with conflictual dialogue as higher in quality or Entertainment than the stories with neutral dialogue. While this may reflect reality in this population, it also seemed plausible that this finding was a Type II error due to methodological flaws, particularly in either the measurement of the dependent variables, or having insufficient power, which might have affected both the t-tests and the multi-level models.

While the PSQI appeared a good option to study story quality and Entertainment, its validity has not been rigorously tested; and the PSQI total score and Entertainment may give insufficiently broad conceptualizations of story quality, and audience response. Moreover, we may have underestimated the sample size needed to detect an effect due to optimistic assumptions about the effect size. Any underestimate would be compounded when the exclusion criteria and imperfect randomization eliminated more participants than predicted (see ESM 2 at Berks, 2026).

These two issues led to the main design changes in Experiment 2.

A second experiment with increased power was felt to be justified to confirm the findings of Experiment 1. Two other major changes were the use of a broader set of outcome measures and placing these measures at the completion of the stories, so participants would rate the stories as a whole.

Table 8 gives descriptive statistics for Experiment 2.

Table 9 gives inferential statistics, indicating that while manipulating the independent variable, Conflict, produces a lower effect size than Experiment 1, this change remains significant. Other dependent variables show similar effects to Experiment 1 — no significant effects (at a p < .05 level) for Quality, Audience Reaction, Enjoyment, or Boredom.

A multi-level model with fixed slopes (see ESM 1 at Berks, 2026) for Quality, ARS Total, Enjoyment, and Boredom, did not show conflict level having a significant effect (at a p < .05 level). In the exploratory analyses, descriptive and inferential statistics (Table 10 and Table 11) of subcategories of the ARS, including Appreciation and Suspense were calculated, showing no significant effects (at a p < .05 level).

Analyses of the interaction between conflict levels and genre/story type are included in ESM 1, showing instances of effects varying across genre/story type.

These experiments investigated whether dramatic conflict, in the form of adversarial dialogue, improved audience ratings of story quality, increased positive emotional/cognitive reactions to a story, or decreased negative reactions.

Making changes to the dialogue, as recommended by the literature, in order to increase dramatic conflict (Patterson, 2013; Stein, 2014) increased the perceived level of conflict in both experiments. The change was statistically significant and had a large to very large effect size (Cohen, 1988). Despite the successful manipulation of conflict, in neither experiment did participants rate the stories with conflictual dialogue more positively over a range of dependent variables. Ostensibly, this is in contrast to experiments showing that people have greater liking for stories with conflict (Brewer & Lichtenstein, 1981; Diener & Woody, 1981; Jose, 1988), and expert opinion that adversarial dialogue is more entertaining (Axelrod, 2013; Bell, 2014; Frey, 2010; Hough, 2015; Stein, 1995).

There are several possible explanations for these findings.

One explanation is that there was no measurable effect because increasing the conflict level in dialogue has no positive effect on audience reaction. Therefore, these results provide no support for the dramatic theory of conflict. There is a parallel in the literature on violence in the media, which has largely failed to show a rise in audience enjoyment with higher levels of violence (Weaver, 2011). However, this lack of positive effect is what would have been predicted by the traditional theory of conflict, which posits that dramatic conflict has a positive effect when it is a consequence of structural conflict. The traditional theory also offers an explanation for the positive effect of conflict in previous experiments which manipulated story conflict (Brewer & Lichtenstein, 1981; Diener & Woody, 1981; Jose, 1988), as it appears likely these were manipulating structural conflict or a combination of structural and dramatic conflict. This null result is also compatible with the psychological theory, which does not have dramatic conflict as a key variable.

Another explanation for the results is that the effect of adversarial conflict was obscured by methodological issues. Potential issues include the strength or dynamics of the manipulation, the studies’ power, and measurement problems.

A stronger manipulation of conflict may have produced a significant response in the dependent variables. Against this, are the response in the manipulation check, and that nothing in the literature suggests a response threshold. More plausibly, conflict that built to a high point and then resolved might have produced more of an effect than the ‘static’ conflict created in these experiments (Egri, 2009; Freytag, 1896). However, characters are expected to act believably (Bates, 1994), and characters engaging in a high level of conflict, either static or building to a peak, which is unrelated to the structural conflict may violate audience expectations.

The study may have been underpowered to detect a very small effect. Despite Experiment 2 having a large sample, it did not achieve the target sample size.

Finally, story-level measures may not be sensitive enough to detect people’s response to dramatic conflict in dialogue. In both studies, participants were asked to rate their reactions to the whole story on the assumption that improvements to dialogue in the dialogue heavy story would translate to better ratings of the story as a whole. If instead, participants were asked to rate a passage of dialogue as a passage, this tighter focus might have allowed detection of meaningful change.

Beyond these limitations, these findings may not generalize. The research tested one form of dramatic conflict, using three different stimuli, in one specific medium (online written stories), consequently, these results only permit strong inferences about the specific stories and medium studied, and cannot necessarily be generalized to a wider “population” of stories and media (Judd et al., 2012; Yarkoni, 2022).

An additional methodological issue, which is unlikely to have affected the result, but could affect its interpretation, is the extent that dramatic conflict was isolated from structural conflict. As there was no check for this isolation in the study design, caution is warranted in interpreting these results as pertaining solely to the effects of dramatic conflict.

The importance of these experiments is in providing practical guidance to writers and helping fill a gap in the current empirical and theoretical literature on the importance of conflict in stories.

While this research is in the form of a psychology experiment, in which a stimulus is manipulated, and people’s responses recorded, it is based on a practical problem. If a passage of dialogue in a story is flat, is it enough for the writer to merely increase the conflict? The results suggest caution against simply adding dramatic conflict to dialogue in the hope of improving story quality and audience reaction. How then should writers approach conflict in dialogue? Firstly, the writer could give any conflict an outcome (something deliberately avoided in these experiments). Previous research into the psychological theory of conflict, strongly suggests that writing dialogue in which a liked protagonist wins a verbal contest with an unliked antagonist will provide enjoyment to an audience, while the reverse outcome will lead to appreciation. This audience response will happen whether or not the adversarial dialogue is strongly linked to structural conflict. Secondly, but more speculatively, there is nothing in these findings which would prevent a writer following the expert opinion in the form of the traditional theory of conflict, where adversarial dialogue springs from the protagonists’ story goals, and obstacles to those goals.

Indeed, the lack of effect on an audience of isolated dramatic conflict, if supported by other research, would turn our attention away from the dramatic theory of conflict to the traditional and psychological theories. Here understanding the effect of dramatic conflict remains important but is only one piece of the puzzle.

This study suggests that future research into conflict could include measuring story quality and audience reaction to different manipulations: manipulating structural conflict while keeping a dramatic scene unchanged; manipulating structural conflict without any dramatic conflict; or manipulating dramatic conflict in a story in which there is a clear valence of character and/or story resolution. Alternatively, research might have a narrower focus, asking participants to rate just a passage of dialogue in which the dramatic conflict has been manipulated. Moving beyond this, the effect of internal vs external conflict, or static conflict versus conflict that builds to a peak could be investigated.

In conclusion, this study examined an under researched part of a broader literature on conflict in stories: whether isolated dramatic conflict, in the form of adversarial dialogue, improves story quality and the audience reaction to the story. No evidence was found supporting this idea. In the absence of similar research, one should not over interpret these results, yet they suggest conflict theories need further experimental testing.