Talker-specificity and adaptation in quantifier interpretation
Introduction
The meaning of many, if not all, words is context-dependent. For example, whether we want to say that John is tall depends on whether John is being compared to other boys his age, professional basketball players, dwarves, etc. (e.g., Halff et al., 1976, Kamp, 1995, Kennedy and McNally, 2005, Klein, 1980). Other words whose interpretation requires reference to context are pronouns and quantifiers (Bach, 2012). For example, the interpretation of a quantifier like many depends on the class of objects that is being quantified over: the number of crumbs that many crumbs refers to is judged to be higher than the number of mountains that many mountains refers to (Hörmann, 1983).
A less-well studied aspect of context-dependence is how a given talker uses quantifiers like many and some. Talkers exhibit individual variability at just about any linguistic level investigated – including, for example, pronunciation (e.g., Allen et al., 2003, Bauer, 1985, Harrington et al., 2000, Yaeger-Dror, 1994), lexical preferences (e.g., Finegan and Biber, 2001, Roland et al., 2007, Tagliamonte and Smith, 2005), and syntactic preferences (e.g., the frequency with which they use passives, Weiner & Labov, 1983). Therefore, talkers are also likely to differ in how they use quantifiers. For example, talkers may differ in how many crumbs they consider to be many crumbs, and these differences would consequently be reflected in their productions. In this case, listeners would be well served by taking into account talker-specific knowledge in order to successfully infer what the talker intended to convey.
Talker-specific knowledge has been observed experimentally in cases of variation in pronunciation and syntactic production (e.g., Clayards et al., 2008, Creel and Bregman, 2011, Creel and Tumlin, 2009, Fine et al., 2013, Kamide, 2012, Kraljic and Samuel, 2007). While this question has received less attention in lexical processing, there is some evidence that listeners can learn to anticipate talker-specific biases in the frequency with which referents are being referred to (Metzing & Brennan, 2003) and that these talker-specific expectations are reflected in online processing (e.g., Creel, Aslin, & Tanenhaus, 2008). These studies complement classic work on conceptual pacts in which interlocutors adjust their use of referential expressions to create temporary, shared context-specific names (Brennan & Clark, 1996).
Previous work on talker-specific lexical expectations has focused on open class, semantically rich, content words – typically nouns (Brennan and Clark, 1996, Creel et al., 2008, Metzing and Brennan, 2003). This raises the question of whether listeners are capable of adapting to talker-specific differences in the use of words that convey more abstract meanings, such as those of quantifiers. If listeners do in fact adapt to talker-specific differences, what specifically are listeners adapting to, i.e., what is the nature of the representations that are being updated and what are the underlying mechanisms?
The current paper begins to address these questions by studying adaptation to talker-specific differences in the use of the quantifiers some and many. We present four experiments that investigate lexical adaptation. Taken together, these experiments establish (i) that listeners can adapt to talker-specific differences in the usage of even abstract lexical items, such as quantifiers; (ii) that, provided sufficient exposure, such adaptation can be achieved even for multiple talkers simultaneously; (iii) that lexical adaptation is observed both to talker-specific differences in the frequency with which lexical items are used and to talker-specific differences in how they are being used; and thus, finally, (iv) that lexical adaptation – although often studied as a separate phenomena—exhibits many of the hallmarks of adaptation observed for other linguistic domains. Next, we elaborate on these points, while introducing the four experiments presented below. In doing so, we relate our research to previous work and highlight the contributions of the current work.
Before we investigate lexical adaptation to talker-specific quantifier use, we first assess whether the premise for adaptation is given: Experiment 1 demonstrates that listeners differ in their initial expectations about a talker’s use of a variety of quantifiers, including some and many. This shows that if listeners want to arrive at an interpretation of an utterance that is close to the talker’s intended meaning, they might sometimes need to adapt their expectations about quantifier use to match those of the current talker. Experiment 1 thus provides the first direct evidence that there would potentially be a benefit to adaptation to talker-specific differences in quantifier use.
This then raises the question whether listeners do adapt to these changes. This is the central motivation for Experiment 2. Going beyond this question and previous work, Experiment 2 also begins to investigate the nature of the changes in expectations that result from exposure to a novel talker. Specifically, we ask whether lexical adaptation can be talker-specific. The answer to this question is of theoretical relevance, as it speaks to the nature of the mechanisms underlying lexical adaptation. We briefly elaborate on this point, as it has so far received relatively little attention in the literature on lexical adaptation (but see Brennan and Clark, 1996, Pickering and Garrod, 2004).
A priori, there are several ways in which a listener can treat experience with a novel talker. A listener might treat new experience as evidence that can be used to sharpen prior expectations about quantifier use without taking into account the specific context, including the talker. Any adaptation would then be to talkers in general. At the other extreme, adaptation might be completely context-specific. If that were the case, then adaptation would be specific to a particular talker in a particular context and would not at all generalize to other talkers. A more likely possibility is that listeners strike a subtle balance between context-general and context-specific adaptation (cf. Kleinschmidt & Jaeger, 2015). Prima facie, it would seem undesirable for a language processing system to allow a small amount of recent exposure to overwrite life-long experience with language. At the same time, it is beneficial to be able to rapidly adapt to talker-specific lexical preferences, potentially increasing the efficiency of communication (for related discussion, see Brennan and Clark, 1996, McCloskey and Cohen, 1989, McRae and Hetherington, 1993, Pickering and Garrod, 2004, Seidenberg, 1994).
One way to meet both the need for adaptation and the need to maintain previously acquired knowledge is to learn and maintain talker-specific expectations, so that adaptation to a novel talker does not imply loss of previously acquired knowledge. Research in speech perception has explored and found support for this hypothesis (Goldinger, 1996, Johnson, 2006, Kraljic and Samuel, 2007; for review, see Kleinschmidt & Jaeger, 2015). More recent research has found support for this idea in other domains of language processing (e.g., prosodic processing, Kurumada et al., 2014, Kurumada et al., 2012; and sentence processing, Fine et al., 2013, Jaeger and Snider, 2013). For example, in episodic and exemplar-based models, linguistics experiences are assumed to be stored along with knowledge about the context in which they occurred (Goldinger, 1996, Johnson, 2006, Pierrehumbert, 2001). This is how these models capture talker-specific expectations. (Similar reasoning applies to Bayesian models of adaptation that assume generative processes over hierarchically organized indexical alignment, Kleinschmidt & Jaeger, 2015). Similarly, memory-based models of lexical alignment (Horton and Gerrig, 2005, Horton and Gerrig, in press) can in theory account for both talker-specific expectations – if talkers are included as contexts (Brown-Schmidt, Yoon, & Ryskin, 2015).
Changes in the use of lexical forms and structures due to exposure are often attributed to temporary changes in expectation within a spreading-activation framework. These “priming-based” accounts assume that exposure increases the activation of a particular word, structure and perhaps conceptually related words and structures (e.g., Arai et al., 2007, Branigan et al., 2005, Chang et al., 2006, Goudbeek and Krahmer, 2012, Pickering and Branigan, 1998, Reitter et al., 2011, Traxler and Tooley, 2008). Although lexical priming accounts have not been applied to the issues we are exploring, the simplest version of these models would most naturally predict that changes in expectations would apply across talkers and thus be talker-independent. In contrast the models discussed above – while compatible with generalization across talkers—predict there to be also talker-specific expectations (as we discuss later, some generalization is, in fact, expected under these alternative accounts).
In Experiment 2a and Experiment 2b, respectively, we ask listeners to either make judgments about “a talker”, which leaves ambiguous the possibility that we are referring to any talker, or “the talker”, referring to the specific talker to whom they were exposed. Comparing the two experiments allow us to ask whether listeners adapt, at least in part, to a specific talker, rather than changing their expectations across the board to reflect how any new talker might use some and many.
Building on the basic effect observed in Experiments 2a and 2b, Experiment 3 then asks whether listeners adjust not only to changes in the frequency with which quantifiers are used by a given talker, but also to changes in how quantifiers are used to refer to specific quantities by a given talker. Both of these quantities are of theoretical interest: talkers might differ in either or both of these aspects, so that the ability to adapt to such differences is potentially beneficial for listeners. Additionally, if lexical adaptation at least qualitatively follows the principles of rational inference and learning (as has been proposed for phonetic adaptation, Kleinschmidt and Jaeger, 2011, Kleinschmidt and Jaeger, 2015, Kleinschmidt and Jaeger, in press and syntactic adaptation, Fine et al., 2010, Kleinschmidt et al., 2012), listeners are expected to be sensitive to both prior probability of quantifiers (i.e., their frequency of use) and the likelihood of quantifiers given an intended interpretation (i.e., how quantifiers are used). Although not framed in these terms, previous work has exclusively focused on adaptation to changes in the frequency (and only for content words, e.g., Creel et al., 2008, Metzing and Brennan, 2003), leaving open whether listeners can adapt to changes in the likelihood. Experiment 3 tests whether listeners can also adapt to changes in the likelihood.
Finally, in Experiment 4 we return to the question of talker-specificity and ask whether listeners can adapt to multiple talkers simultaneously, when these talkers differ in how they use some and many. This prediction is made by episodic (Goldinger, 1996), exemplar-based (Johnson, 1997, Johnson, 2006, Pierrehumbert, 2001), and certain Bayesian models (Kleinschmidt & Jaeger, 2015) of adaptation in speech perception. Talker-specific adaptation to multiple talkers has been observed in experiments on speech perception (Kraljic & Samuel, 2007) and, more recently, during syntactic processing (Kamide, 2012). To the best of our knowledge, it has not previously been tested for lexical processing. Experiment 4 exposes listeners to two talkers with different usage of some and many.
The studies presented here thus extends previous research on lexical adaptation and alignment in comprehension both methodologically – by establishing the exposure-test paradigm frequently used in research on speech perception as suitable for research on lexical adaptation—and empirically. We find that listeners can adapt to both how often and with what intended interpretation specific talkers use some and many, and that – at least in simple situations like those investigated here—listeners can adapt to talker-specific quantifier use of multiple talkers from very little input. The experiments presented here establish a novel paradigm to investigate lexical adaptation in ways parallel to research on adaptation to talker variability in speech perception. This makes our results comparable to research in these other fields. Indeed, we find several parallels between lexical adaptation and adaptation at other levels of language processing. We close by discussing venues for future research on lexical adaptation that, we think, are facilitated by the current paradigm.
Section snippets
Experiment 1: Variability in quantifier interpretation
It is well-known that there are gradient context-dependent differences in the interpretation of quantifiers (e.g., Hörmann, 1983, Newstead, 1988, Pepper and Prytclak, 1974). It is less clear, however, whether talkers differ in their use of quantifiers. For example, talkers could differ in the overall frequency with which they use a certain quantifier, in their interpretation of a quantifier (i.e., when they will use it), or both. If there is such variation, different listeners – who have been
Experiment 2: Adaptation of beliefs about quantifier use based on recent input
Experiment 2 investigates whether listeners can adjust their beliefs about the use of some and many based on recent input specific to the current context. We used a variation of the exposure-and-test paradigm frequently used in research on perceptual learning, including research on speech perception (e.g., Eisner and McQueen, 2006, Kraljic and Samuel, 2007, Norris et al., 2003, Vroomen et al., 2007). A post-exposure test assessed participants’ beliefs about the typical use of some and many.
Experiment 3: Adapting to the frequency vs. use of quantifiers
Experiment 3 employed the identical procedure as Experiment 2b, with only one change: participants were exposed to an equal number of many and some trials. Specifically, the exposure talker produced one of the quantifiers in its prototypical usage (based on Experiment 1 results and confirmed below). For the other quantifier, the exposure talker had the same ‘biased’ usage employed in Experiment 2. We describe this manipulation in more detail below.
By equating the frequency of many and some
Experiment 4: Adapting to multiple talkers
Taken together, Experiments 2 and 3 suggest that exposure to relatively few trials is sufficient for listeners to adapt their expectations about how a given talker uses many and some – at least, when the talker is observed producing highly informative descriptions as in the current experiments, where the talker is observed producing multiple critical trials describing the same domain.
As we noted earlier, the need for adaptation and the need to maintain previously acquired knowledge can be
General discussion
The studies reported in this paper used a web-based paradigm to explore a specific type of context-dependence that has received comparatively little attention in the literature, talker-specific differences in how quantifiers are used. In particular we focused on adaptation to talker-specific use of some and many, drawing parallels to recent work on adaptation in other domains of language, with a special focus on phonetic adaptation—the domain, which has been most widely investigated to date.
Conclusion
The experiments reported in this paper suggests that even minimal exposure to a speaker whose use of quantifiers differs from a listener’s expectations can result in a talker-specific shift in that listener’s beliefs about future quantifier use. Our results further suggest that listeners adapt to both the frequency with which a talker uses certain words and the specific interpretation intended by the talker. This complements work on adaptation in other domains, for example, adaptation in
Acknowledgments
We thank Andrew Watts for technical support. We thank the editor and the four anonymous reviewers for their helpful comments on the earlier versions of this manuscript. Parts of this study were presented at the 35th Annual Conference of the Cognitive Science Society, the 20th Architectures and Mechanisms for Language Processing, and at Discourse Expectations: Theoretical, Experimental, and Computational Perspectives 2015. This work was partially supported by a post-doctoral fellowship to IY
References (71)
- et al.
Priming ditransitive structures in comprehension
Cognitive Psychology
(2007) - et al.
Mixed-effects modeling with crossed random effects for subjects and items
Journal of Memory and Language
(2008) Tracing phonetic change in the received pronunciation of British English
Journal of Phonetics
(1985)- et al.
Chapter three – People as contexts in conversation
Psychology of Learning and Motivation
(2015) - et al.
Consistency in interpretation of probabilistic phrases
Organizational Behavior and Human Decision Processes
(1985) - et al.
Perception of speech reflects optimal use of probabilistic speech cues
Cognition
(2008) - et al.
Heeding the voice of experience: The role of talker variation in lexical access
Cognition
(2008) - et al.
Would a blue kite by any other name be just as blue? Effects of descriptive choices on subsequent referential behavior
Journal of Memory and Language
(2014) - et al.
The impact of memory demands on audience design during language production
Cognition
(2005) - et al.
Alignment as a consequence of expectation adaptation: Syntactic priming is affected by the prime’s prediction error given both prior and recent experience
Cognition
(2013)
Resonance in an exemplar-based lexicon: The emergence of social identity and phonology
Journal of Phonetics
Learning individual talkers’ structural preferences
Cognition
Prototype theory and compositionality
Cognition
Perceptual adjustments to multiple speakers
Journal of Memory and Language
When conceptual pacts are broken: Partner-specific effects on the comprehension of referring expressions
Journal of Memory and Language
Quantifiers as fuzzy concepts
Perceptual learning in speech
Cognitive Psychology
Sometimes frequently means seldom: Context effects in the interpretation of quantitative expressions
Journal of Research in Personality
The representation of verbs: Evidence from syntactic priming in language production
Journal of Memory and Language
Frequency of basic English grammatical structures: A corpus analysis
Journal of Memory and Language
Language and connectionism: The developing interface
Cognition
Visual recalibration and selective adaptation in auditory–visual speech perception: Contrasting build-up courses
Neuropsychologia
Individual talker differences in voice-onset-time
Journal of the Acoustical Society of America
Saying, meaning, and implicating
Priming prepositional-phrase attachment during comprehension
Journal of Experimental Psychology: Learning, Memory, and Cognition
Conceptual pacts and lexical choice in conversation
Journal of Experimental Psychology: Learning, Memory, and Cognition
Becoming syntactic
Psychological Review
Whatever next? Predictive brains, situated agents, and the future of Cognitive Science
Behavioral and Brain Sciences
How talker identity relates to language processing
Language and Linguistics Compass
Perceptual learning in speech: Stability over time
The Journal of the Acoustical Society of America
Evidence for implicit learning in syntactic comprehension
Cognitive Science
Rapid expectation adaptation during syntactic comprehension
PloS One
Is there syntactic adaptation in language comprehension?
Cited by (40)
Priming reveals similarities and differences between three purported cases of implicature: Some, number and free choice disjunctions
2021, Journal of Memory and LanguageI know what you're probably going to say: Listener adaptation to variable use of uncertainty expressions
2020, CognitionCitation Excerpt :However, since all our experiments had a between-subjects design, it could be that participants were only adapting to the experimental situation, independent of the speaker. This seems unlikely given the results reported by Yildirim et al. (2016), who found that participants formed speaker-specific production expectations after being exposed to multiple speakers whose use of quantifiers differed. Moreover, Schuster and Degen (2019) have provided evidence of speaker-specific adaptation to uncertainty expressions.
Repeat after us: Syntactic alignment is not partner-specific
2019, Journal of Memory and LanguageCounting ‘uhm's: How tracking the distribution of native and non-native disfluencies influences online language comprehension
2019, Journal of Memory and LanguageCitation Excerpt :Emerging frameworks in psycholinguistics (e.g., P-chain model: Dell & Chang, 2013; Belief-updating model: Kleinschmidt & Jaeger, 2015) state that listeners overcome the variability and uncertainty that is so pervasive in spontaneous spoken communication by flexibly adjusting their expectations about future input based on prior experiences. Distributional learning within the scope of a single experimental session has been demonstrated in the online processing of prosodic (Kurumada, Brown, & Tanenhaus, 2012), phonological (Norris, McQueen, & Cutler, 2003), syntactic (Fine, Jaeger, Farmer, & Qian, 2013), and semantic cues (Yildirim, Degen, Tanenhaus, & Jaeger, 2016) in language. The current study contributes to these psycholinguistic perspectives by showing, for the first time, that listeners also track the distribution of metalinguistic performance cues, namely disfluencies like uh.
Genre and linguistic expectation shift: Evidence from pop song lyrics
2019, Language in Society