Computational Interpretations of the Gricean Maxims in the Generation of Referring Expressions

We examine the problem of generating definite noun phrases that are appropriate referring expressions; that is, noun phrases that (a) successfully identify the intended referent to the hearer whilst (b) not conveying to him or her any false conversational implicatures (Grice, 1975). We review several possible computational interpretations of the conversational implicature maxims, with different computational costs, and argue that the simplest may be the best, because it seems to be closest to what human speakers do. We describe our recommended algorithm in detail, along with a specification of the resources a host system must provide in order to make use of the algorithm, and an implementation used in the natural language generation component of the IDAS system.     

Utterance Interpretation and Actual Intentions

Axiomathes - In this paper I argue, from the consideration of what I hope is the complete variety of a hearer’s approaches to a speaker’s utterance, that (1) the speaker’s...     

Local pragmatics in a Gricean framework

The pragmatic framework developed by H.P. Grice in “Logic and Conversation” explains how a speaker can mean something more than, or different from, the conventional meaning of the sentence she utters. But it has been argued that the framework cannot give a similar explanation for cases where these pragmatic effects impact the understood content of an embedded clause, such as the antecedent of a conditional, a clausal disjunct, or the clausal complement of a verb. In this paper, I show that such an explanation is available. One of the central arguments of the paper (Section 2) is that in a significant subset of cases, local pragmatic effects are a consequence of a global (utterance-level) pragmatic requirement. In these cases, local pragmatic effects are a consequence of ‘acting locally’ to resolve a potential global pragmatic violation. These cases do not require us to posit application of pragmatic principles (Maxims of Conversation) to the contents of embedded clauses. The account does, though, require the assumption that interpreters can identify and reason about the contents of unasserted sub-parts of sentences, an assumption that I motivate in section 3. Building on this, in section 4 of the paper, I argue that once we have recognized that interpreters can, and do, reason independently about the contents of non-asserted clauses, it becomes unproblematic to assume that in some cases, Gricean conversational principles do apply directly to these contents, providing an alternative route to account for local pragmatic effects. In revisiting the ideas of this paper in my response to the commentaries, I consider in more detail the revisions to Grice’s broader program that are necessitated by these moves, in particular acknowledging the problematicity of Grice’s notion of what is said. I argue that the starting point for Gricean reconstructions should instead be merely what is expressed, which carries no pragmatic commitments regarding what is speaker meant.     

Testimonial Knowledge and Transmission

We often talk about knowledge being transmitted via testimony. This suggests a picture of testimony with striking similarities to memory. For instance, it is often assumed that neither is a generative source of knowledge: while the former transmits knowledge from one speaker to another, the latter preserves beliefs from one time to another. These considerations give rise to a stronger and a weaker thesis regarding the transmission of testimonial knowledge. The stronger thesis is that each speaker in a chain of testimonial transmission must know that p in order to pass this knowledge to a hearer. The weaker thesis is that at least the first speaker must know that p in order for any hearer in the chain to come to know that p via testimony. I argue that both theses are false, and hence testimony, unlike memory, can be a generative source of knowledge.     

Testimonial Knowledge from Lies

Recently, Dan O’Brien has argued that there are situations in which a hearer can gain testimonial knowledge from a speaker who is lying. In order to make his case, O’Brien presents two examples where a speaker lies to a hearer, but supposedly the hearer comes to have testimonial knowledge on the basis of the lying speaker’s testimony. O’Brien claims that his examples demonstrate that lies can be used to pass on knowledge in a non-inferential fashion. I argue that O’Brien is mistaken. More specifically, I argue that the hearer’s belief in the second example that O’Brien depicts fails to meet two plausible conditions for knowledge. First, the hearer’s belief fails to satisfy the requirements of the epistemic basing relation. Second, the hearer’s belief is not safe.     

Evidential Preemption

As a general rule, whenever a hearer is justified in forming the belief that p on the basis of a speaker’s testimony, she will also be justified in assuming that the speaker has formed her belief appropriately in light of a relevantly large and representative sample of the evidence that bears on p. In simpler terms, a justification for taking someone’s testimony entails a justification for trusting her assessment of the evidence. This introduces the possibility of what I will call “evidential preemption.” Evidential preemption occurs when a speaker, in addition to offering testimony that p, also warns the hearer of the likelihood that she will subsequently be confronted with apparently contrary evidence: this is done, however, not so as to encourage the hearer to temper her confidence in p in anticipation of that evidence, but rather to suggest that the (apparently) contrary evidence is in fact misleading evidence or evidence that has already been taken into account. Either way, the speaker is signalling to the hearer that the subsequent disclosure of this evidence will not require her to significantly revise her belief that p. Such preemption can effectively inoculate an audience against future contrary evidence, and thereby creates an opening for a form of exploitative manipulation that I will call “epistemic grooming.” Nonetheless, I argue, not all uses of evidential preemption are nefarious; it can also serve as an important tool for guiding epistemically limited agents though complex evidential scenarios.     

Relevance theory,pragmatic inference and cognitive architecture

Relevance Theory (RT) argues that human language comprehension processes tend to maximize “relevance,” and postulates that there is a relevance-based procedure that a hearer follows when trying to understand an utterance. Despite being highly influential, RT has been criticized for its failure to explain how speaker-related information, either the speaker’s abilities or her/his preferences, is incorporated into the hearer’s inferential, pragmatic process. An alternative proposal is that speaker-related information gains prominence due to representation of the speaker within higher level goal-directed schemata. Yet the goal-based account is still unable to explain clearly how cross-domain information, for example linguistic meaning and speaker-related knowledge, is integrated within a modular system. On the basis of RT’s cognitive requirements, together with contemporary cognitive theory, we argue that this integration is realized by utilizing working memory and that there exist conversational constraints with which the constructed utterance interpretation should be consistent. We illustrate our arguments with a computational implementation of the proposed processes within a general cognitive architecture.

Abbreviations: ACT-R Adaptive Control of Thought - RationalCOGENT Cognitive Objects within a Graphical ENvironmenTCS/SS Contention Scheduling/Supervisory SystemRBCP Relevance-Based Comprehension ProcedureRT Relevance Theory     

Local pragmatics: reply to Mandy Simons

In response to Mandy Simons’ defence of a classical Gricean approach to pragmatic enrichment in terms of conversational implicature, I emphasize the following contrast. Conversational implicatures are generated by a global inference which uses as a premise the fact that the speaker has said that p, but only the triggering inference is global in cases of pragmatic enrichment. What generates the correct interpretation is a process of reconstrual, which locally maps the literal meaning of a constituent to a modulated meaning and composes that meaning with that of the other constituents. That process is constrained by Gricean considerations (in the broad sense) but that is true of all pragmatic aspects of interpretation, whether pre-propositional or post-propositional. Just as indexical resolution, though pragmatic and constrained by Gricean considerations, does not fit the two-stage model through which Grice accounts for conversational implicatures, so pragmatic modulation can’t be accounted for in terms of that model despite the fact that, like conversational implicatures and unlike indexical resolution, modulation is pragmatically rather than semantically triggered.     

The impact of politeness and relationship on perceived quality of advice about a problem

Advice is a common but potentially problematic way to respond to someone who is distressed. Politeness theory (Brown & Levinson, 1987) suggests advice threatens a hearer's face and predicts that the speaker‐hearer relationship and the use of politeness strategies can mitigate face threat and enhance the effectiveness of advice messages. Students (N=384) read 1 of 16 hypothetical situations that varied in speaker power and closeness of the speaker‐hearer relationship. Students then read 1 of 48 advice messages representing different politeness strategies and rated the message for regard shown for face and for effectiveness. However, neither speaker‐hearer relationship nor politeness strategies was consistently associated with perceived threat to face or perceived advice effectiveness. We suggest revisions to politeness theory and additional factors that may affect judgments of face sensitivity and advice effectiveness.     

Presupposition Projection and Conditionalization

I explain what exactly constrains presupposition projection in compound sentences and argue that the presuppositions that do not project are conditionalized, giving rise to inferable conditional presuppositions. I combine elements of (Gazdar in Pragmatics: implicature, presupposition, and logical form. Academic Press, New York 1979) and (van der Sandt in Context and presupposition. Croom Helm, London 1988) which, together with an additional, independently motivated assumption, make it possible to construct an analysis that makes correct predictions. The core of my proposal is as follows: When a speaker felicitously utters a compound sentence whose constituent clauses (considered in isolation) require presuppositions, the hearer will infer that the speaker presupposes those propositions, unless the sentence contains some element that makes the hearer realize that, if the speaker actually presupposed them, she would be either uninformative or inconsistent in her beliefs. In these cases, the propositions that would have been presupposed, had the clauses been uttered in isolation, will not be presupposed, i.e. the clausal presuppositions will not project.     

Telling as Inviting to Trust*

How can I give you a reason to believe what I tell you? I can influence the evidence available to you. Or I can simply invite your trust. These two ways of giving reasons work very differently. When a speaker tells her hearer that p, I argue, she intends that he gain access to a prima facie reason to believe that p that derives not from evidence but from his mere understanding of her act. Unlike mere assertions, acts of telling give reasons directly. They give reasons by inviting the hearer's trust. This yields a novel form of anti‐reductionism in the epistemology of testimony. The status of testimony as a sui generis source of epistemic warrant is entailed by the nature of the act of telling. We can discover the nature of this illocution, and its epistemic role, by examining how it functions in the real world of human relations.     

Persuasive Paradoxes in Cicero's Speeches

The paper first presents a short survey of ancient and modern logical, rhetorical and argumentative approaches (e.g. Aristotle, Quintilian, Quine, Anscombre and Ducrot) studying the properties of paradoxical utterances. This survey is followed by a tentative definition of paradoxes as seemingly contradictory utterances triggering conversational implicatures in the sense of Grice. A specific group of paradoxes, namely, persuasive paradoxes, is further characterized by the specific implicatures which they trigger: the implicatures of persuasive paradoxes serve the interest of the (political) speaker because they either convey a sharp criticism of the political opponent(s) or praise the political activities of the speaker in a highly effective way.The second part of the paper takes a corpus of about 80 paradoxical utterances from Cicero's speeches to show how they are used 1. for a devastating criticism of Cicero's political enemies, 2. a milder form of criticism in the case of his friends, when their political activities have failed, 3. a praise of successful policies of Cicero and his political friends and 4. a defense of unsuccessful activities started by Cicero and his friends.     

Evaluating the multiple proposition strategy

Contextualism about many expressions faces a common objection: in some discourses it appears that there is no single interpretation which can explain how a speaker is justified in making her assertion and how a hearer with different information or standards is justified in negatively evaluating what the speaker said. According to the Multiple Proposition Strategy, contextualists may attempt to explain these competing features pragmatically in terms of different propositions in play. In this paper I argue against the Multiple Proposition Strategy, first focusing on epistemic modals and then generalising the results to other expressions. I show how when purportedly contextualist terms are embedded in belief reports, we get similar problems but that the Multiple Proposition Strategy does not provide a satisfactory explanation of such cases. I suggest, therefore, that we reject the Multiple Proposition Strategy in favour of a theory that explains the unembedded and embedded cases in similar ways.     

Promise is debt, threat another matter: the effect of credibility on the interpretation of conditional promises and threats.

Evidence from reasoning tasks shows that promises and threats both tend to receive biconditional interpretations. They also both display high speaker control. On the face of it, the only difference seems to be the positive or negative signing of the consequent. In a promise, the speaker tries to persuade the hearer to do something by holding out the prospect of a particular reward; in a threat, the speaker tries to refrain the hearer from doing something by holding out the prospect of a particular punishment. This paper investigates the respects in which conditional promises and threats differ further by means of an inference task. The credibility of the consequent was manipulated in order to examine whether the acceptability ratings of inferences based on promises and on threats would be equally affected. The results of the inference task and an analysis of the reasons people give for their answers suggest that the credibility of promises is less affected by the use of excessive consequents than the credibility of threats. In other words, promise remains debt, whereas threat is another matter.     

Semantic minimalism and the “miracle of communication”

According to semantic minimalism, context-invariant minimal semantic propositions play an essential role in linguistic communication. This claim is key to minimalists’ argument against semantic contextualism: if there were no such minimal semantic propositions, and semantic content varied widely with shifts in context, then it would be “miraculous” if communication were ever to occur. This paper offers a critical examination of the minimalist account of communication, focusing on a series of examples where communication occurs without a minimal semantic proposition shared between speaker and hearer. The only way for minimalists to respond to these examples is by restricting the scope of their account to intra-lingual communication. It can then be shown (1) that the minimalist’s notion of a language shrinks to a point, such that practically no instances of communication will fall under that account, and (2) that the retreat to intra-lingual communication is in any case self-defeating, since the only way for minimalists to account for the individuation of languages is by resort to precisely the kinds of contextual considerations they abjured in the first place. In short, if, as minimalists allege, contextualism founders because it renders communication contingent on speaker and hearer sharing a context, it can now be seen that minimalism faces a parallel problem because it renders communication contingent on speaker and hearer sharing a language. I end by arguing that the possibility of communication cannot, as minimalists assume, be grounded in shared semantic conventions; rather, successful communication must precede the establishment of any particular set of semantic conventions.     

Cross-Cultural Comparisons of the Perceived Importance of Conversational Constraints

This study investigates how cultural groups may differ in their perceptions about preferred communication behavior. It focuses on the importance attached to “conversational constraints” in conversation. For the primary goal of requesting, the following five conversational constraints are identified: (a) concern for clarity, (b) concern for avoiding hurting the hearer's feelings, (c) concern for nonimposition, (d) concern for avoiding negative evaluation by the hearer, and (e) concern for effectiveness. Five hypotheses are formulated regarding the relationships between individualistic-collectivistic orientation (an important dimension of“culture”) and these five conversational constraints. Data were collected from 892 participants (undergraduates studying in Korea, Hawaii, and the mainland United States). After being presented at random with one of the six request situations, participants rated the perceived importance of each conversational constraint in that situation. The results indicate that the perceived importance of clarity was higher in the more individualistic cultures. On the other hand, the perceived importance of avoiding hurting the hearer's feelings and of minimizing imposition were higher in the more collectivistic cultures. The perceived importance of effectiveness and of avoiding negative evaluation by the hearer, on the other hand, did not significantly differ across the three cultural groups. Theoretical as well as practical implications are discussed.     

The effects of psychological gender orientations on the perceived salience of conversational constraints

This study investigated psychological gender differences in the salience of conversational constraints. It was hypothesized that feminine self-orientations would correlate positively with concern for the hearer’s feelings and concern for avoiding negative evaluation by the hearer, and that masculine self-orientations would correlate positively with concern for clarity. Furthermore, we hypothesized that of the various gender-orientation patterns, androgynous individuals (maintaining high masculine as well as high feminine self-identity simultaneously) will have the highest importance ratings of all the constraints added together; i.e., higher than either undifferentiated (low masculine as well as low feminine self-identity simultaneously) or gender-typed individuals (either low masculine-high feminine or high masculine-low feminine). Partkipants consisted of undergraduate students from diverse ethnic backgrounds. After being presented with four conversational situations, participants rated the perceived importance of each constraint in relation to each situation. They then completed the Bem’sSex- Role Inventory to measure their psychological gender orientation. The results were mostly consistent with the hypotheses. Furthermore, the study showed that one’s psychological gender-role orientation is superior to biological sex categories in accounting for the perceived importance of conversational constraints.     

Resistance to the influences of others: Limits to the formation of a collective memory through conversational remembering

People often form collective memories by sharing their memories with others. Warnings about the reliability of one conversational participant can limit the extent to which conversations or other forms of postevent information can influence subsequent memory. Although this attenuation is consistently found for prewarnings, there are substantial reasons to suspect that, by carefully manipulating both individual characteristics of the listener in a conversation and the dynamics of the postevent conversation, one can restrict the effect even prewarnings have on the influence a speaker might have on the memory of a listener. Indeed, in situations in which a speaker contributes substantially to a conversation and the quality of memory of a listener is poor, prewarnings have the paradoxical effect of increasing the influence of the speaker on a listener's memory. Warnings may not always limit the formation of a collective memory. Copyright © 2009 John Wiley & Sons, Ltd.     

The Concept of Trying

It is widely held that whenever someone φs, that person tries to do φ. I examine arguments by B. O'Shaughnessy and J. Hornsby, and considerations by P. Grice in support of that thesis. I argue that none of them are convincing. The remainder of the paper defends an analysis of the concept of trying along the lines opposed by Grice et al. By speaking of someone's trying to φ the speaker leaves the room for failure or the possibility of failure. This account of the function of the word 'try' makes it clear what is wrong with the claim that whenever someone φs, that person tries to do φ. For that claim amounts to saying that every action is either a failure or a possible failure – which clearly one can't say of any action that one has already acknowledged to be successful. Furthermore, whether an action is possibly a failure depends on how much one knows about it. Hence the question whether in itself an action does or does not involve trying is as misguided as the question whether it's intrinsically unexpected. It may be correct for some to describe it as 'an attempt to φ', but incorrect for others.