Abstract
One of the major open issues in semantics and pragmatics concerns the role of convention in relating sentence types with illocutionary acts and perlocutionary effects. For the type-to-illocution connection, some degree of force conventionalism seems to be widely accepted. In contrast, Austin (1962) and many subsequent researchers have assumed that perlocution is not a matter of convention, but rather arises inexorably from illocution, content, and context. In this paper, we challenge this fundamental assumption about perlocution with evidence from a new perception experiment focused on perlocutionary effects relating to the listener’s conception of the speaker as a social actor. We find that these effects are predictable from sentence type plus intonation (‘type + tune’), that they vary by type + tune, and that they are consistent across a wide range of sentence contents, contexts, and illocutionary inferences. We argue that these conventions are naturally incorporated into existing work on sentence-type conventions.
Introduction: Challenging the Perlocutionary Assumption
In classical speech-act theory, a distinction is drawn between illocution (the act performed in speaking, such as requesting) and perlocution (the consequential effect by speaking, such as offending or appearing polite). While sentence types (declarative, interrogative, etc.) are widely recognized to have conventional ties to illocutionary acts, perlocutionary effects have traditionally been viewed as non-conventional, unpredictable byproducts of context and content.
This research challenges that foundational assumption. By investigating how listeners perceive social traits like annoyance, authority, and politeness, we demonstrate that specific combinations of sentence type and intonation ("type + tune") reliably trigger consistent social perceptions. These findings suggest that languages possess irreducible conventions for perlocutionary effects, independent of the specific illocutionary act being performed.
Methodology
To test the conventionality of perlocutionary effects, we conducted perception experiments with 240 native American English speakers. We utilized a custom dataset of 31 sentences across four types: polar interrogatives, wh-interrogatives, declaratives, and imperatives.
To isolate the effect of intonation, sentences were recorded with a flat terminal contour and then synthetically manipulated using the PSOLA algorithm in Praat. This created three distinct, highly controlled terminal contours for every utterance: rising, level, and falling. Participants rated these clips on two dimensions: the intended illocutionary act and perlocutionary social traits (annoyance, authority, and politeness) on a 0–100 scale.
Key Insights and Results
The "Big Three" Tunes
The experiments revealed a remarkably stable mapping between intonation and social perception across all sentence types:
- Falling contour: Consistently associated with higher perceived authority.
- Level contour: Reliably signals annoyance.
- Rising contour: Correlates with politeness and a positive social stance.
Illocutionary Independence
A critical finding is that these perlocutionary effects remain stable even when controlling for the perceived illocutionary act. While intonation can shift whether a listener hears a sentence as a request or an invitation (as seen in the varying counts in Figure 1), the social judgment of the speaker's attitude (e.g., sounding annoyed) persists regardless of which act the listener identifies.
Type + Tune Interactions
While the core effects are stable, sentence types and tunes interact to modulate intensity. For instance, a rising declarative ("uptalk") reduces perceived authority more significantly than a rising interrogative, likely because it violates the canonical falling tune associated with declaratives. Conversely, declaratives with a level contour are perceived as significantly less annoyed than other sentence types using the same contour.
Conclusion
Our results indicate that perlocutionary effects are not merely context-driven accidents but are governed by predictable linguistic conventions. These "type + tune" conventions allow speakers to reliably project social personas and emotional states. This suggests that the strict boundary between illocutionary and perlocutionary conventions should be reconsidered in favor of a model where both are integrated into the conventional meaning of sentence types and intonational contours.
Summary updated