Accent modulates access to word meaning: Evidence for a speaker-model account of spoken word recognition

Speech carries accent information relevant to determining the speaker’s linguistic and social background. A series of web-based experiments demonstrate that accent cues can modulate access to word meaning. In Experiments 1–3, British participants were more likely to retrieve the American dominant meaning (e.g., hat meaning of “bonnet”) in a word association task if they heard the words in an American than a British accent. In addition, results from a speeded semantic decision task (Experiment 4) and sentence comprehension task (Experiment 5) confirm that accent modulates on-line meaning retrieval such that comprehension of ambiguous words is easier when the relevant word meaning is dominant in the speaker’s dialect. Critically, neutral-accent speech items, created by morphing British- and American-accented recordings, were interpreted in a similar way to accented words when embedded in a context of accented words (Experiment 2). This finding indicates that listeners do not use accent to guide meaning retrieval on a word-by-word basis; instead they use accent information to determine the dialectic identity of a speaker and then use their experience of that dialect to guide meaning access for all words spoken by that person. These results motivate a speaker-model account of spoken word recognition in which comprehenders determine key characteristics of their interlocutor and use this knowledge to guide word meaning access.     

The power of the written word: Comparing word meaning inferencing from auditory and written modalities throughout grade school

Previous research indicates that word learning from auditory contexts may be more effective than written context at least through fourth grade. However, no study has examined contextual differences in word learning in older school-aged children when reading abilities are more developed. Here we examined developmental differences in children’s ability to deduce the meanings of unknown words from the surrounding linguistic context in the auditory and written modalities and sought to identify the most important predictors of success in each modality. A total of 89 children aged 8–15 years were randomly assigned to either read or listen to a narrative that included eight novel words, with five exposures to each novel word. They then completed three posttests to assess word meaning inferencing. Children across all ages performed better in the written modality. Vocabulary was the only significant predictor of success on the word inferencing task. Results indicate support for written stimuli as the most effective modality for novel word meaning deduction. Our findings suggest that the presence of orthographic information facilitates novel word learning even for early, less proficient readers.     

Propose but verify: Fast mapping meets cross-situational word learning

We report three eyetracking experiments that examine the learning procedure used by adults as they pair novel words and visually presented referents over a sequence of referentially ambiguous trials. Successful learning under such conditions has been argued to be the product of a learning procedure in which participants provisionally pair each novel word with several possible referents and use a statistical-associative learning mechanism to gradually converge on a single mapping across learning instances [e.g., Yu, C., & Smith, L. B. (2007). Rapid word learning under uncertainty via cross-situational statistics. Psychological Science, 18(5), 414–420]. We argue here that successful learning in this setting is instead the product of a one-trial procedure in which a single hypothesized word-referent pairing is retained across learning instances, abandoned only if the subsequent instance fails to confirm the pairing – more a ‘fast mapping’ procedure than a gradual statistical one. We provide experimental evidence for this propose-but-verify learning procedure via three experiments in which adult participants attempted to learn the meanings of nonce words cross-situationally under varying degrees of referential uncertainty. The findings, using both explicit (referent selection) and implicit (eye movement) measures, show that even in these artificial learning contexts, which are far simpler than those encountered by a language learner in a natural environment, participants do not retain multiple meaning hypotheses across learning instances. As we discuss, these findings challenge ‘gradualist’ accounts of word learning and are consistent with the known rapid course of vocabulary learning in a first language.     

No frills: Simple regularities in language can go a long way in the development of word knowledge

Recent years have seen a flourishing of Natural Language Processing models that can mimic many aspects of human language fluency. These models harness a simple, decades-old idea: It is possible to learn a lot about word meanings just from exposure to language, because words similar in meaning are used in language in similar ways. The successes of these models raise the intriguing possibility that exposure to word use in language also shapes the word knowledge that children amass during development. However, this possibility is strongly challenged by the fact that models use language input and learning mechanisms that may be unavailable to children. Across three studies, we found that unrealistically complex input and learning mechanisms are unnecessary. Instead, simple regularities of word use in children's language input that they have the capacity to learn can foster knowledge about word meanings. Thus, exposure to language may play a simple but powerful role in children's growing word knowledge. A video abstract of this article can be viewed at https://youtu.be/dT83dmMffnM .

Research Highlights

• Natural Language Processing (NLP) models can learn that words are similar in meaning from higher-order statistical regularities of word use.
• Unlike NLP models, infants and children may primarily learn only simple co-occurrences between words.
• We show that infants' and children's language input is rich in simple co-occurrence that can support learning similarities in meaning between words.
• We find that simple co-occurrences can explain infants' and children's knowledge that words are similar in meaning.

     

Conventions’ Revenge: Davidson,Derangement, and Dormativity

Davidson advocates a radical and powerful form of anti-conventionalism, on which the scope of a semantic theory is restricted to the most local of contexts: a particular utterance by a particular speaker. I argue that this hyper-localism undercuts the explanatory grounds for his assumption that semantic meaning is systematic, which is central, among other things, to his holism. More importantly, it threatens to undercut the distinction between word meaning and speaker’s meaning, which he takes to be essential to semantics. I argue that a moderate form of conventionalism can restore systematicity and the word/speaker distinction while accommodating Davidson’s insights about the complexities and contextual variability of language use.     

Measuring incremental changes in word knowledge: Experimental validation and implications for learning and assessment

The goal of this study was to test a new technique for assessing vocabulary development. This technique is based on an algorithm for scoring the accuracy of word definitions using a continuous scale (Collins-Thompson & Callan, 2007). In an experiment with adult learners, target words were presented in six different sentence contexts, and the number of informative versus misleading contexts was systematically manipulated. Participants generated a target definition after each sentence, and the definition-scoring algorithm was used to assess the degree of accuracy on each trial. We observed incremental improvements in definition accuracy across trials. Moreover, learning curves were sensitive to the proportion of misleading contexts, the use of spaced versus massed practice, and individual differences, demonstrating the utility of this procedure for capturing specific experimental effects on the trajectory of word learning. We discuss the implications of these results for measurement of meaning, vocabulary assessment, and instructional design.     

Concreteness and word production

Two experiments are reported that investigated the effect of concreteness on the ability to generate words to fit sentence contexts. When participants attempted to retrieve words from dictionary definitions in Experiment 1, abstract words were associated with more omissions and more alternates than were concrete words. These findings are consistent with the view that the semantic–lexical weights in the word production system are weaker for abstract than for concrete words. We found no evidence that greater competition from semantic neighbors was an additional reason why abstract words were harder to produce. Participants also reported more positive tip-of-the-tongue states (TOTs) when attempting to produce abstract words from their definitions, consistent with more phonological retrieval problems for abstract than for concrete words. In Experiment 2, participants attempted to generate words to fit into a sentence that described a specific event. The difference between the numbers of abstract and concrete words recalled was significantly smaller in the event condition than in the definition condition, and evidence no longer emerged of greater phonological retrieval failure for abstract words. Overall, the results are consistent with the view that the semantic–lexical weights, but not the lexical–phonological weights, are weaker for abstract than for concrete words in the word production system.     

Word Senses as Clusters of Meaning Modulations: A Computational Model of Polysemy

Most words in natural languages are polysemous; that is, they have related but different meanings in different contexts. This one-to-many mapping of form to meaning presents a challenge to understanding how word meanings are learned, represented, and processed. Previous work has focused on solutions in which multiple static semantic representations are linked to a single word form, which fails to capture important generalizations about how polysemous words are used; in particular, the graded nature of polysemous senses, and the flexibility and regularity of polysemy use. We provide a novel view of how polysemous words are represented and processed, focusing on how meaning is modulated by context. Our theory is implemented within a recurrent neural network that learns distributional information through exposure to a large and representative corpus of English. Clusters of meaning emerge from how the model processes individual word forms. In keeping with distributional theories of semantics, we suggest word meanings are generalized from contexts of different word tokens, with polysemy emerging as multiple clusters of contextually modulated meanings. We validate our results against a human-annotated corpus of polysemy focusing on the gradedness, flexibility, and regularity of polysemous sense individuation, as well as behavioral findings of offline sense relatedness ratings and online sentence processing. The results provide novel insights into how polysemy emerges from contextual processing of word meaning from both a theoretical and computational point of view.     

The role of gaze direction and mutual exclusivity in guiding 24‐month‐olds' word mappings

In these studies, we examined how a default assumption about word meaning, the mutual exclusivity assumption and an intentional cue, gaze direction, interacted to guide 24‐month‐olds' object‐word mappings. In Expt 1, when the experimenter's gaze was consistent with the mutual exclusivity assumption, novel word mappings were facilitated. When the experimenter's eye‐gaze was in conflict with the mutual exclusivity cue, children demonstrated a tendency to rely on the mutual exclusivity assumption rather than follow the experimenter's gaze to map the label to the object. In Expt 2, children relied on the experimenter's gaze direction to successfully map both a first label to a novel object and a second label to a familiar object. Moreover, infants mapped second labels to familiar objects to the same degree that they mapped first labels to novel objects. These findings are discussed with regard to children's use of convergent and divergent cues in indirect word mapping contexts.     

Early word meanings: The case of object names

It has been claimed that young children use object names overgenerally and undergenerally because they do not have notions of objects of particular kinds, but rather, complexive notions of objects and their habitual actions or locations. However, for overgeneral uses in particular, it is difficult to differentiate word meaning from word use because communicative functions are not explicitly expressed in single-word speech. In the present paper, we identify three types of overgeneral uses, and argue that two of these reflect communicative functions rather than complexive meanings. We obtained production data from 10 children in the single-word period, using a standardized method of recording utterance contexts. Most uses of object names were for appropriate instances of the adult categories. Of the overgeneral uses, most were attributable to communicative functions rather than complexive meanings, and there was no evidence of undergeneral use. The results provide strong evidence that, from the very start, children's object names, like those of adults, apply to objects of particular kinds.     

Context change and the role of meaning in word recognition

Two experiments were conducted in which the central variable was the addition of verbal context on the test which was not present during study. Neither study showed context to be a relevant variable. Because verbal context effects (when they do occur) are presumed to result from a change of meaning, the broader role of word meaning in recognition was examined. Studies in which homographs were used to produce a change in meaning were reviewed with the conclusion that when appropriate controls are used the effects are too small to support meaning as a major factor underlying recognition. An experiment using homographs verified the general conclusion from previous studies. A review of studies was then undertaken to determine the influence of synonym distractors on recognition scores. If word meaning is critical in word recognition, the use of synonyms of study words as new words on the recognition test should produce large decrements. The evidence available did not support this expectation, and an experiment using both within-subject and between-subject comparisons showed at best only small effects. The general conclusion was that theories of recognition in which word meaning plays a dominant role cannot be supported. It is likely that word meaning is encoded during study of lists of words but is infrequently used in making recognition decisions.     

Semantic diversity: A measure of semantic ambiguity based on variability in the contextual usage of words

Semantic ambiguity is typically measured by summing the number of senses or dictionary definitions that a word has. Such measures are somewhat subjective and may not adequately capture the full extent of variation in word meaning, particularly for polysemous words that can be used in many different ways, with subtle shifts in meaning. Here, we describe an alternative, computationally derived measure of ambiguity based on the proposal that the meanings of words vary continuously as a function of their contexts. On this view, words that appear in a wide range of contexts on diverse topics are more variable in meaning than those that appear in a restricted set of similar contexts. To quantify this variation, we performed latent semantic analysis on a large text corpus to estimate the semantic similarities of different linguistic contexts. From these estimates, we calculated the degree to which the different contexts associated with a given word vary in their meanings. We term this quantity a word’s semantic diversity (SemD). We suggest that this approach provides an objective way of quantifying the subtle, context-dependent variations in word meaning that are often present in language. We demonstrate that SemD is correlated with other measures of ambiguity and contextual variability, as well as with frequency and imageability. We also show that SemD is a strong predictor of performance in semantic judgments in healthy individuals and in patients with semantic deficits, accounting for unique variance beyond that of other predictors. SemD values for over 30,000 English words are provided as supplementary materials.     

Learning to represent word meaning: What initial training events reveal about children's developing action verb concepts

Very little experimental research has shown how the events from which verbs are learned affect children's representation of meaning. This study addressed these lacunae by systematically exploring how three different initial training contexts affect children's and adults' interpretation of novel action verbs. Brief, videotaped action events were used to teach children and adults novel verbs in 1 of 3 conditions. Subjects were asked whether these verbs generalized to other events differing only in outcome, manner, instrument, or agent. Initial representation of verb meaning was inferred from generalizations. Unlike 10-year-olds and adults, 3-year-olds' interpretations were (a) significantly context specific—preferring instrument and outcome elements of meaning in one context, but manner in another, and (b) consistently, though moderately biased to favor the manner of action. Results clearly demonstrate that both the young word learner and the context in which word learning occurs are equally important determinants of developing action verb concepts.     

中性抽象词和具体词情绪信息获得的差异

采用词汇迫选学习范式考察不同类型词汇获得情绪信息的认知机制。结果发现：（1）在迫选过程中与积极图片匹配的具体词汇喜好度显著高于与消极图片匹配的喜好度,而抽象词学习并未表现出这样的效应。（2）积极图片下被选择词汇的喜好度显著高于未被选择词汇的喜好度,而消极图片下被选择词汇与未被选择词汇的喜好度差异不显著。该结果表明具体词相较于抽象词更易获得相应情绪信息,积极情绪比消极情绪信息更易被习得。     

The role of partial knowledge in statistical word learning

A critical question about the nature of human learning is whether it is an all-or-none or a gradual, accumulative process. Associative and statistical theories of word learning rely critically on the later assumption: that the process of learning a word’s meaning unfolds over time. That is, learning the correct referent for a word involves the accumulation of partial knowledge across multiple instances. Some theories also make an even stronger claim: Partial knowledge of one word–object mapping can speed up the acquisition of other word–object mappings. We present three experiments that test and verify these claims by exposing learners to two consecutive blocks of cross-situational learning, in which half of the words and objects in the second block were those that participants failed to learn in Block 1. In line with an accumulative account, Re-exposure to these mis-mapped items accelerated the acquisition of both previously experienced mappings and wholly new word–object mappings. But how does partial knowledge of some words speed the acquisition of others? We consider two hypotheses. First, partial knowledge of a word could reduce the amount of information required for it to reach threshold, and the supra-threshold mapping could subsequently aid in the acquisition of new mappings. Alternatively, partial knowledge of a word’s meaning could be useful for disambiguating the meanings of other words even before the threshold of learning is reached. We construct and compare computational models embodying each of these hypotheses and show that the latter provides a better explanation of the empirical data.     

Sensitivity to sampling in Bayesian word learning

We report a new study testing our proposal that word learning may be best explained as an approximate form of Bayesian inference (Xu & Tenenbaum, in press). Children are capable of learning word meanings across a wide range of communicative contexts. In different contexts, learners may encounter different sampling processes generating the examples of word-object pairings they observe. An ideal Bayesian word learner could take into account these differences in the sampling process and adjust his/her inferences about word meaning accordingly. We tested how children and adults learned words for novel object kinds in two sampling contexts, in which the objects to be labeled were sampled either by a knowledgeable teacher or by the learners themselves. Both adults and children generalized more conservatively in the former context; that is, they restricted the label to just those objects most similar to the labeled examples when the exemplars were chosen by a knowledgeable teacher, but not when chosen by the learners themselves. We discuss how this result follows naturally from a Bayesian analysis, but not from other statistical approaches such as associative word-learning models.     

Motor processing modulates word comprehension

We evaluated whether movement modulates the semantic processing of words. To this end, we used homograph words with two meanings, one associated with hand movements (e.g., ‘abanico’, ‘fan’ in Spanish) or foot movements (‘bota’, ‘boot’ in Spanish), and the other not associated with movement (‘abanico’, ‘range’ in Spanish; ‘bota’, ‘wineskin’ in Spanish). After the homograph, three words were presented, and participants were asked to choose the word related to one of the two homograph meanings. The words could be either related to the motor meaning of the homograph (‘fan-heat’), to the non-motor meaning of the homograph (‘range-possibility’) or unrelated (‘fan-phone’). The task was performed without movement (simple condition) or by performing hand (Experiment 1) and foot (Experiment 2) movements. Compared with the simple condition, the performance of movement oriented the preference towards the motor meaning of the homograph. This pattern of results confirms that movement modulates word comprehension.     

Interplay of emotional valence and concreteness in word processing: An event-related potential study with verbs

The functional locus of emotional valence in word processing remains an open question. In event-related potentials, emotion has been found to elicit an early posterior negativity (EPN), which is assumed to reflect attention catching by the words’ meaning. Previously, the EPN was modulated by word category with verbs exhibiting longer EPN latencies compared with other word categories. Here we examined whether concreteness, a semantic variable, influences emotion processing. Within a lexical decision task for verbs, emotional valence (positive, negative, and neutral) was orthogonally combined with concreteness (concrete and abstract). EPN onset was found already at 250 ms post-stimulus for concrete verbs, whereas it started 50 ms later for abstract verbs. Concreteness effects occurred after the start of main effects of emotion. Thus, the elicitation of the EPN seems to be based on semantic processes, with emotional valence being accessed before other semantic aspects such as concreteness of verbs.     

Predictive coding accelerates word recognition and learning in the early stages of language development

The ability to predict future events in the environment and learn from them is a fundamental component of adaptive behavior across species. Here we propose that inferring predictions facilitates speech processing and word learning in the early stages of language development. Twelve‐ and 24‐month olds’ electrophysiological brain responses to heard syllables are faster and more robust when the preceding word context predicts the ending of a familiar word. For unfamiliar, novel word forms, however, word‐expectancy violation generates a prediction error response, the strength of which significantly correlates with children's vocabulary scores at 12 months. These results suggest that predictive coding may accelerate word recognition and support early learning of novel words, including not only the learning of heard word forms but also their mapping to meanings. Prediction error may mediate learning via attention, since infants’ attention allocation to the entire learning situation in natural environments could account for the link between prediction error and the understanding of word meanings. On the whole, the present results on predictive coding support the view that principles of brain function reported across domains in humans and non‐human animals apply to language and its development in the infant brain. A video abstract of this article can be viewed at: http://hy.fi/unitube/video/e1cbb495-41d8-462e-8660-0864a1abd02c . [Correction added on 27 January 2017, after first online publication: The video abstract link was added.]     

The developing role of prosody in novel word interpretation

This study examined whether children use prosodic correlates to word meaning when interpreting novel words. For example, do children infer that a word spoken in a deep, slow, loud voice refers to something larger than a word spoken in a high, fast, quiet voice? Participants were 4- and 5-year-olds who viewed picture pairs that varied along a single dimension (e.g., big vs. small flower) and heard a recorded voice asking them, for example, “Can you get the blicket one?” spoken with either meaningful or neutral prosody. The 4-year-olds failed to map prosodic cues to their corresponding meaning, whereas the 5-year-olds succeeded (Experiment 1). However, 4-year-olds successfully mapped prosodic cues to word meaning following a training phase that reinforced children’s attention to prosodic information (Experiment 2). These studies constitute the first empirical demonstration that young children are able to use prosody-to-meaning correlates as a cue to novel word interpretation.