Elsevier

Cognition

Volume 118, Issue 2, February 2011, Pages 234-244
Cognition

What’s new? Children prefer novelty in referent selection

https://doi.org/10.1016/j.cognition.2010.10.015Get rights and content

Abstract

Determining the referent of a novel name is a critical task for young language learners. The majority of studies on children’s referent selection focus on manipulating the sources of information (linguistic, contextual and pragmatic) that children can use to solve the referent mapping problem. Here, we take a step back and explore how children’s endogenous biases towards novelty and their own familiarity with novel objects influence their performance in such a task. We familiarized 2-year-old children with previously novel objects. Then, on novel name referent selection trials children were asked to select the referent from three novel objects: two previously seen and one completely novel object. Children demonstrated a clear bias to select the most novel object. A second experiment controls for pragmatic responding and replicates this finding. We conclude, therefore, that children’s referent selection is biased by previous exposure and children’s endogenous bias to novelty.

Introduction

The problem of learning words is a problem of ambiguity wrapped in uncertainty and shrouded in vagueness. On the auditory side, the acoustic cues to individual phonemes are inherently context dependent, and at the level of words, there are few cues that allow the listener to unambiguously segment a word. On the visual and conceptual side, parsing objects from the visual scene is extraordinarily difficult, and how the properties of those objects are mapped to concepts and categories is still not understood. Children, of course, must solve both problems without substantial top-down knowledge, and without always knowing all the relevant features of the input.

However, learning a word requires children to face a third level of ambiguity: they must learn which of those highly ambiguous auditory categories maps to which of those highly uncertain visual categories. This mapping problem can be seen quite vividly the first time a child is presented with a novel word and several possible referents. In such situations, there are an infinite number of interpretations for the novel word that might be considered (cf., Quine, 1960)—each of the referents, their properties, actions that the speaker may take on them (or has taken) and so forth. Each is instantiated as a possible link between the auditory word-form and concept, and thus, even in this relatively constrained task, an enormous number of potential mappings must be considered. The rapid pace at which young children typically acquire a vocabulary, however, points to their skill at solving this problem.

Historically, research aimed at examining this skill has focused on the types of information children use to determine the correct referent of a novel word. In a typical study, the child is presented with a novel word, some possible referents for that word, and various cues such as eye-gaze, gestures, or syntactic frames that suggest the intended meaning. Such studies have been incredibly productive and reveal, among other things, the contributions of linguistic, conceptual and social/pragmatic cues in children’s determination of novel word meaning (see e.g., Baldwin, 1991, Booth and Waxman, 2003, Heibeck and Markman, 1987, Jaswal and Markman, 2001, Moore et al., 1999, Tomasello and Akhtar, 1995). The main focus of this kind of work, however, is how children use the linguistic or pragmatic cues presented by the experimenter in conjunction with a novel word and the available objects to determine the referent. As such, this approach treats early word learning as an inference problem and the cues presented in the task as sources of information that can be used to constrain the problem-space.

Studies based on this approach attempt to minimize the influence of individual children’s endogenous biases on their in-the-moment performance by using completely novel objects and/or words, focusing on mean performance across groups of children, and controlling the experiment such that only one or two sources of information are available. Importantly, however, children’s endogenous biases could create an uneven playing field on which other sources of information must compete. As an example, consider a case in which social cues such as eye-gaze direct the child to one object, but the child is endogenously biased to consider another object. If a novel word were presented, which object would win and be mapped to the novel word? The one the child was biased toward? Or the one the other speaker was indicating with eye-gaze? Evidence from Hirsh-Pasek, Golinkoff, and Hollich (2000) suggests that there is a developmental progression in word learning tasks that pit object salience against other cues to referents. Likewise, there are similar changes in the role of object salience in directing eye-gaze during the second year (see, Moore (2008) for a review and Corkum and Moore, 1998, Deák et al., 2000).

Clearly, any complete account of children’s behavior in even the simplest of word learning tasks needs to take children’s biases, goals and intentions into account. Biases may not always align with the task and/or empirical questions posed by the experimenter, and it is crucial to be aware of such biases so that they can be accounted for. But, more importantly, such biases may compete with information-based processes to determine how the child treats novel words and novel objects in the real world as well. The studies presented here take a first step in this direction by examining an endogenous bias that is quite likely to influence children’s performance in referent selection tasks and can be manipulated in the laboratory—children’s attraction to novel objects.

There is considerable evidence from early infancy that novelty plays a role in driving attention, eye-movements and other behaviors, and that this may change with development (e.g., Hunter & Ames, 1988). For example, from 2 months of age infants decrease their amount of looking to familiar stimuli across successive trials, as the stimuli become more familiar (Fantz, 1964), from 4 months infants discriminate between previously seen targets and previously unseen foils from the same category after less than 15 s of exposure (e.g., Quinn, Eimas, & Rosenkrantz, 1993), and from 7 months infants reach more for novel objects than familiar objects (Shinskey & Munakata, 2005). Thus, there is fairly robust evidence for a basic novelty bias in how infants approach stimuli.

It is likely that children’s relative familiarity with objects and words would also bias their performance in the referent selection tasks that occur every day. For example, in the 17,000 words that each child hears on a given day (Hart & Risley, 1995), there are likely to be many word forms that may be partly or entirely familiar, but for which the child does not know the meaning. Similarly, a child who knows 50, 100 or even 200 words is likely surrounded every day by thousands of objects for which she has not yet learned the names, but that she has seen several times before. Thus, while every word and object is novel to a child at one point in time, outside of laboratory experiments, the proportion of situations in which children are presented with words and objects with which they have no familiarity whatsoever is likely to be quite small.

Previous research has made it clear that children can use the novelty of potential referents as a cue to word meaning. For example, a child presented with an object from a category for which she knows the name, such as a cup, and an object from a novel, nameless category will map the novel name to the novel object (Halberda, 2006). Or, if only the cup is present, the child will infer the novel name refers to an object part such as the handle (Hansen & Markman, 2009). Children can also use relative novelty in a context to determine if a word denotes an object or action (Tomasello & Akhtar, 1995). In highly ambiguous cases, in which an adult does not use a name but requests “it,” 12–14-month-old infants will choose the object they have never played with before, that is, the object that is most novel to themselves, but by 18–20 months infants are just as likely to choose the object that is most novel to the adult (MacPherson & Moore, 2010). Similarly, after an adult and infant have played with an object together, 14–20-month-old infants will map the ambiguous pronoun “it” to something previously played with, rather than something novel (Ganea & Saylor, 2007). Thus, novelty in context contributes to children’s ability to handle referential ambiguity. However, children are not simply novelty-seekers—in these experiments, novelty contributes as a source of information along with other pragmatic, lexical, and memory-based processes.

Other studies suggest that children take novelty relative to a speaker into account in referent selection. For example, children take adults’ excitement as an indication of what is most novel and is therefore more likely to be the referent of a novel word (Tomasello & Haberl, 2003). Likewise, in cases where they do not know names for any of the stimuli but the novel object is familiar to the experimenter, children map novel names to novel parts (Grassmann et al., 2009, Moll et al., 2007). Along this line, previous research by Akhtar and colleagues (Akhtar et al., 1996, Diesendruck et al., 2004) suggests that children use the relative novelty of an object to a speaker when mapping a novel name to a collection of nameless objects (see also, Moll and Tomasello, 2007, Moll et al., 2007). On the basis of such data, these authors and others (see e.g., Birch & Bloom, 2002) have argued that novelty acts as a social pragmatic cue in naming contexts.

Both lines of research argue that novelty to either the speaker or the child participates as a source of information in complex inference processes—children become increasingly sensitive to what other people know and novelty can serve as a component of that sensitivity. But can novelty play a more basic role? A number of researchers have argued that endogenous attentional mechanisms, removed from communicative pragmatics, can bias children’s interpretations of novel words. For example, Samuelson and Smith (1998) found that children readily map novel names to the most salient object in a context; suggesting that simple endogenous mechanisms could shape a lexical response quite independently of the demands of word learning. In previous studies, because the use of novelty as a cue to word meaning emphasized the inferential component of referent selection, the literature neglected the contribution of children’s own endogenous biases toward or away from novel objects. Samuelson and Smith argued that such salience may even be a proxy or heuristic for other types of inference—that is, social/pragmatic cues may work, at least early in development, by biasing attention. However, whether such endogenous biases are equivalent to more inferential processes, or compete against them to shape the response, it is clear that we must understand them in order to understand word learning in its natural environment. Samuelson and Smith’s (1998) data offer a good start and a hint that such biases may exist, however, their study focused on short-term attentional factors—the increased salience of an object based on its novelty in a context—not longer-developing factors like the familiarity of an object to a child. This the focus of the work presented here.

To better understand how children’s relative familiarity with objects, and their endogenous biases towards novelty, influence their interpretation of novel names, we need a systematic investigation of a novel word learning situation in which no information is presented to bias referent selection. A particularly informative case would be one in which children are confronted with a novel word and multiple novel, nameless objects. Lacking any additional information, children should be equally likely to select any of the possible referents. But what if the objects varied in how familiar they were to the child? All of the objects are nameless, so again, each object should be equally likely to be selected as the referent of the novel name. However, the extent to which children are biased to select objects based on their novelty/familiarity would reveal something about the internally driven biases that shape this behavior.

To this end, we tested the effects of children’s familiarity with objects on how they map novel names to novel objects. Specifically, we presented 2-year-old children with a referent selection task in which the only difference between the possible referents was the child’s own previous experience with the objects. Thus, in the current study we ask if children’s own familiarity with the novel, nameless objects changes how they perform on a referent selection task. Importantly, when the completely novel object is introduced in this study there is no change in location (cf., Diesendruck et al., 2004), no change in the experimenter’s excitement (cf., Akhtar et al., 1996), no special activity enacted only on it (cf., Samuelson & Smith, 1998) and no information provided about whether it is familiar or novel to the experimenter (cf., Birch & Bloom, 2002). That is, we do not highlight novelty in any way and do not use novelty as a cue to a “correct” answer. If children still map the novel name onto the most novel object, this will suggest that children’s endogenous biases to novelty play an important role in young children’s referent selections.

Section snippets

Participants

Twelve 24-month-old monolingual, English-speaking children (M = 24m 20d, SD = 14.11d, range = 24m 7d – 25m 20d, 5 girls, 7 boys) participated. Data from one additional child were excluded from analyses due to experimenter error. Parents were reimbursed for parking and children received a small gift (e.g., storybook) for participating.

Stimuli

Sixteen known and 16 unknown objects served as potential referents (see Fig. 1). Before the experiment began, the experimenter showed the parent color photographs of the

Participants

Twelve 24-month-old monolingual, English-speaking children (M = 24m 26d, SD = 19.2, range = 23m 24d – 26m 4d, 6 girls, 6 boys) participated. None of the children had participated in Experiment 1. Data from two additional children were excluded from the analyses due to parental interference (n = 1), which included naming the stimuli and picking objects for the child, and fussiness (n = 1).

Design and procedure

All aspects of the experiment were identical to Experiment 1 except that one experimenter presented objects during the

General discussion

Across two experiments we familiarized children with previously novel objects and then presented them with referent selection trials with a novel name, two now-familiar novel objects and a completely new “supernovel” object. Children behaved systematically in this task: they mapped novel names to the object that was most novel to themselves. This was the case even if all the objects were equally novel to the adult speaker at the time of referent selection (Experiment 2). This study differed

Acknowledgements

This research was supported by NICHD Grant 5R01HD045713 to LKS. Portions of these data were presented at the 2009 Biennial Meeting of the Society for Research in Child Development (Denver, USA) and New Directions in Word Learning (York, UK). We thank Richard Aslin for comments on an earlier version and Ryan Schiffer, Amanda Falk and the students in the Iowa Language and Category Development Laboratory for assistance in data collection. We also thank the parents and children who participated.

References (50)

  • A.E. Booth et al.

    Mapping words to the world in infancy: Infants’ expectations for count nouns and adjectives

    Journal of Cognition and Development

    (2003)
  • V. Corkum et al.

    The origins of joint visual attention in infants

    Developmental Psychology

    (1998)
  • G. Deák et al.

    Effects of gesture and target on 12- and 18-month-olds’ joint visual attention to objects in front of or behind them

    Developmental Psychology

    (2000)
  • G. Diesendruck et al.

    Two-year-olds’ sensitivity to speakers’ intent: An alternative account of Samuelson and Smith

    Developmental Science

    (2004)
  • A.E. Ellis et al.

    Infants flexibly use different dimensions to categorize objects

    Developmental Psychology

    (2006)
  • R.L. Fantz

    Visual experience in infants: Decreased attention familar patterns relative to novel ones

    Science

    (1964)
  • P. Floor et al.

    Can 18-month-old infants learn words by listening in on conversations?

    Infancy

    (2006)
  • P. Ganea et al.

    Infants’ use of shared linguistic information to clarify ambiguous requests

    Child Development

    (2007)
  • R.M. Golinkoff et al.

    Young-children and adults use lexical principles to learn new nouns

    Developmental Psychology

    (1992)
  • M.B. Hansen et al.

    Children’s use of mutual exclusivity to learn labels for parts of objects

    Developmental Psychology

    (2009)
  • B. Hart et al.

    Meaningful differences in the everyday experience of young American children

    (1995)
  • T.H. Heibeck et al.

    Word learning in children – An examination of fast mapping

    Child Development

    (1987)
  • K. Hirsh-Pasek et al.

    An emergentist coalition model for word learning: Mapping words to objects is a product of the interaction of multiple cues

  • Horst, J. S., McMurray, B., & Samuelson, L. K. (2006). Online processing is essential for leaning: Understanding fast...
  • J.S. Horst et al.

    Fast mapping but poor retention by 24-month-old infants

    Infancy

    (2008)
  • Cited by (82)

    • Twelve-month-olds disambiguate new words using mutual-exclusivity inferences

      2021, Cognition
      Citation Excerpt :

      Heuristics to match novel words and unfamiliar objects would lead to increased orienting to unfamiliar objects (relative to the infants' baseline looking behaviour) irrespective of what happened before labelling. While such heuristics might contribute to disambiguation via mutual exclusivity in some experimental contexts (and mostly in older children, Horst et al., 2011; Mather & Plunkett, 2012), they cannot explain the current results. Since the age group tested here was younger than the children who were shown to use novelty-matching strategies, it is possible that such strategies require more language experience to develop (see also Lewis et al., 2020).

    • Sometimes it is better to know less: How known words influence referent selection and retention in 18- to 24-month-old children

      2020, Journal of Experimental Child Psychology
      Citation Excerpt :

      The fact that children also selected the novel item when it was a foil in known RS also fits this idea. Thus, these data coincide with prior work showing a role for novelty in referent selection (Horst, Samuelson, Kucker, & McMurray, 2011; Kucker, McMurray, & Samuelson, 2018; Mervis & Bertrand, 1994). However, although more novelty is known to support selection, retention from novelty-based referent selection is less well documented.

    View all citing articles on Scopus
    View full text