The relationships among verbal short-term memory, phonological awareness, and new word learning: Evidence from typical development and Down syndrome

Abstract

This study examined the correlates of new word learning in a sample of 64 typically developing children between 5 and 8 years of age and a group of 22 teenagers and young adults with Down syndrome. Verbal short-term memory and phonological awareness skills were assessed to determine whether learning new words involved accurately representing phonological information in memory. Results showed a relationship between verbal short-term memory measures and typically developing individuals’ ability to learn the phonological form of novel words but not their ability to learn the physical referent of new words. Similarly, individuals with Down syndrome showed impaired verbal short-term memory and impaired form but not referent learning. Together, these findings specify the circumstances in which an accurate phonological representation within short-term memory is required for new word learning.

Introduction

Verbal short-term memory is typically assessed by asking participants to repeat a sequence of words in correct order as soon as these words have been presented. Although such tasks are extremely simple in form, they are the focus of considerable research interest because of the suggestion (e.g., Baddeley, Gathercole, & Papagno, 1998) that verbal short-term memory capacity constrains individuals’ new word learning skills. Baddeley and colleagues’ (1998) theoretical argument is that a dedicated verbal short-term memory system has evolved in humans to support new word learning because in order to create a long-term representation of a novel word, one must first maintain it in short-term memory (see also Gathercole, 2006). Other computational models are consistent with the suggestion that the formation of stable, long-term phonological representations depends on the quality of the short-term phonological representations of words that are created when they are first encountered (see Burgess and Hitch, 2005, Gupta, 2003, Gupta and MacWhinney, 1997).

Empirical evidence also indicates an association between verbal short-term memory performance and new word learning (see Gathercole, 2006). Individuals with apparently strong verbal short-term memory skills tend to have a more extensive receptive vocabulary, a finding that has been reported both in children learning their native language (Bowey, 1996, Gathercole et al., 1999, Gathercole et al., 1992, Majerus, Poncelet, Greffe, et al., 2006; see also Baddeley et al., 1998) and in individuals learning second languages (e.g., Masoura and Gathercole, 1999, Papagno and Vallar, 1995, Service and Kohonen, 1995). More directly, studies of experimental new word learning in children have shown a relationship between measures of individuals’ verbal short-term memory performance and their ability to learn the new names of novel objects or characters (e.g., Gathercole et al., 1997, Jarrold et al., 2004, Michas and Henry, 1994, Mosse and Jarrold, 2008; see also Gupta, 2003).

Although these findings suggest that verbal short-term memory capacity plays a causal role in aspects of vocabulary acquisition as Baddeley, Gathercole, and colleagues would suggest, others have argued that this association is in fact mediated by other factors. A number of authors have argued that as individuals’ receptive vocabulary increases, their phonological awareness develops via a process of “lexical restructuring” (Bowey, 2001, Metsala, 1999, Metsala and Walley, 1998). Phonological awareness is the understanding of, and the ability to process and manipulate, the component speech sounds of one’s language. The claim is that phonological representations become more precisely specified and distinct as a direct consequence of increased receptive vocabulary knowledge because representing lexical entries in terms of a limited set of common phonemes is more efficient than representing the phonological structure of each word separately (Metsala & Walley, 1998). Proponents of this view argue that phonological awareness plays a role in determining performance on experimental tests of both verbal short-term memory and new word learning because of the need to accurately encode and maintain phonological information in such tasks (Bowey, 1996, de Jong et al., 2000, Messbauer and de Jong, 2003, Snowling et al., 1991, Windfuhr and Snowling, 2001). Therefore, it is possible that, as a direct consequence of vocabulary development, developing phonological awareness mediates the apparent relationship between verbal short-term memory and new word learning (Bowey, 1996). Importantly, however, the potential interrelations between these constructs means that determining their causal priority in predicting new word learning is not straightforward (Service, 2006).

Down syndrome is one condition that can shed important light on the potential correlates of novel word learning. Caused by triplication of the 21st chromosome, trisomy 21 (LeJeune, Gautier, & Turpin, 1959), Down syndrome is associated with particular deficits both in verbal short-term memory performance (e.g., Brock and Jarrold, 2005, Laws, 2002, Numminen et al., 2001; for a review, see Jarrold, Purser, & Brock, 2006) and in phonological awareness (Cossu et al., 1993, Gombert, 2002, Roch and Jarrold, 2008, Snowling et al., 2002, Verucci et al., 2006), and so word learning difficulties would certainly be expected in this condition (Abbeduto et al., 2007, Chapman et al., 2006, Gathercole and Alloway, 2006, Kay-Raining Bird et al., 2004).

Preliminary support for the suggestion that a deficit in verbal short-term memory performance in Down syndrome is causally related to problems in word learning comes from a study by Laws and Gunn (2004) in which the verbal short-term memory performance of a group of individuals with Down syndrome at one time point was found to predict their vocabulary level at a second time point 5 years later. However, in general, vocabulary skills in Down syndrome appear not to be as impaired as might be predicted on the basis of short-term memory skills. Receptive vocabulary is consistently observed as a relative strength among the various aspects of language function in Down syndrome (Chapman, 1995, Chapman et al., 1991), and although it is often delayed relative to age levels, it tends to be superior to individuals’ syntactic skills, for example (Laws and Bishop, 2003, Miller, 1988, Vicari et al., 2002).

One explanation for this apparently puzzling finding is that performance on receptive vocabulary tasks may well be relatively insensitive to the quality of any given phonological representation held in verbal short-term memory. The fact that the experimenter reprovides the phonological label in such tasks means that individuals need only to match this sound to the phonological representation they themselves have maintained. This matching may be perfectly achievable even if the maintained phonological representation is substantially degraded (Chapman et al., 2006, Kay-Raining Bird and Chapman, 1998). In fact, one would expect verbal short memory to particularly relate to the production of novel words given that one clearly cannot produce a more accurate utterance than one can maintain in memory (Kay-Raining Bird et al., 2004; cf. Gathercole et al., 1997, Gupta et al., 2006, Service et al., 2007).

The production of new word forms has been examined in Down syndrome in the context of “fast mapping” studies. Fast mapping is the ability to pair a novel word form with a novel object (e.g., Carey and Bartlett, 1978, Mervis and Bertrand, 1995) and is typically tested in an incidental learning paradigm where the experimenter refers to the novel object in a general linguistic context. At test, participants’ knowledge of the new word form is assessed either in a “comprehension” task, where they need to select the appropriate object when the experimenter names it, or in a “production” task, where they are asked to name the novel object. Although some studies have shown evidence of problems in fast mapping in Down syndrome (Chapman, 2003, Kay-Raining Bird et al., 2000), this is not observed consistently (Chapman et al., 1990, Chapman et al., 2006). However, all of these previous studies credited participants with a correct productive response when they successfully produced at least two of the three phonemes in the novel consonant–vowel–consonant (CVC) label. This no doubt reflects the fact that problems in speech production are common in Down syndrome (e.g., Dodd & Thompson, 2001), and this in turn makes it difficult to assess the integrity of individuals’ phonological representations in productive tasks, particularly those employing nonword stimuli (Cairns and Jarrold, 2005, Laws, 1998). Nevertheless, the failure of these studies to penalize individuals who produced imperfect responses means that they do not provide a stringent assessment of the quality of the novel phonological representation on which these responses are based. Therefore, these existing studies say less than they might about the potential associations among verbal short-term memory, phonological awareness, and new word learning.

The current study aimed to provide a direct test of the quality of phonological representations acquired during new word learning and of the situations in which an accurate phonological representation is a prerequisite for successful acquisition. This was done by assessing word learning in two ways and by presenting these tasks to typically developing children and to individuals with Down syndrome who are known to have relative problems in the representation and maintenance of phonological information. First, in a referent learning task, participants were asked to select the object that had previously been paired with a given phonological form in a manner analogous to a comprehension test of a fast mapping study. Second, in a form learning task that paralleled the production aspect of the fast mapping procedure, participants were required to identify the phonological label that had previously been associated with a given object. To circumvent any speech production problems, rather than participants being required to produce this label themselves, they needed to select the appropriate item from among three choices that were provided to them auditorily (cf. Storkel, 2001). By ensuring that the target and foil sounds were phonologically similar to each other, we emphasized the need for an accurate phonological representation, as would be the case in a stringent production task. A key prediction of the study was that the factors that affect individuals’ ability to maintain such a phonological representation, notably verbal short-term memory capacity but perhaps also phonological awareness, would be related to success on form learning. In contrast, given the arguments outlined above, we predicted that these factors would not heavily influence performance in the referent learning task because a partial phonological representation would be sufficient to mediate a match between the heard word form and the correct object.

One other manipulation was made to these word learning tasks to examine another potential influence on word learning performance, namely, the extent to which individuals employ existing receptive vocabulary knowledge to support their new word learning (cf. Baddeley, 1993, Gathercole, 2006). Individuals’ knowledge of the language does constrain their memory performance (Hulme et al., 1997, Roodenrys et al., 2002, Thorn and Gathercole, 2001, Thorn et al., 2005), and this is true even when nonword stimuli are employed in tests of verbal short-term memory, with more “wordlike” nonwords being easier to recall than less wordlike ones (e.g., Gathercole et al., 1999, Roodenrys and Hinton, 2002, Thorn and Frankish, 2005). Beneficial effects of the “wordlikeness” of to-be-learned nonwords have also been observed on new word learning in both children and adults (Storkel, 2001, Storkel et al., 2006, Storkel and Rogers, 2000). The wordlikeness of nonwords can be defined in at least two ways: at the lexical level and at the sublexical level. At the lexical level, nonwords can share phonemes with “near neighbors” within the participant’s existing lexicon. At the sublexical level, nonwords can consist of phoneme combinations (typically measured as biphones) that are more or less common in the native language. Although these measures of neighborhood size and biphone frequency tend to be highly related to one another (Bailey & Hahn, 2001), they can be separated and evidence suggests that both may be factors that affect the memorability of nonword stimuli (Thorn and Frankish, 2005, Thorn and Jarrold, 2007). In the current study, we investigated these effects on word learning, and the related issue of whether individuals with Down syndrome rely on these linguistic influences to an atypical extent, by varying both the biphone frequency and neighborhood size of the nonword sets that we employed.

In sum, the broad aim of the current study was to determine the different factors that affect the different components of new word learning. Specifically, typically developing children and individuals with Down syndrome were presented with two tests of word learning—form learning and referent learning—with each containing conditions that systematically varied the wordlikeness of the nonwords to be learned. To assess the potential correlates of these two aspects of word learning among typically developing individuals, and possible causes of word learning problems for individuals with Down syndrome, measures of verbal short-term memory and phonological awareness were included. Three verbal short-term memory tasks were employed. These was a standard serial recall task in which participants were required to immediately repeat lists of words that were presented to them auditorily. To further examine potential group differences in degree of linguistic support to verbal short-term memory performance, this task had two conditions that varied the number of lexical neighbors of the word sets.¹ Two recognition tests that examined verbal short-term memory for either item or order information were also employed (cf. Brock and Jarrold, 2004, Majerus, Ponceelt, Elsen, et al., 2006, Majerus, Poncelet, Greffe, et al., 2006). Of the three measures of phonological awareness that were presented to participants, two (rhyme judgment and alliteration detection) had been used previously by Snowling and colleagues (2002) in their study of phonological awareness skills in Down syndrome. The third measure was a test of initial phoneme deletion (cf. Muter, Hulme, & Snowling, 1997).

Individuals’ speech perception skills were also measured using a minimal pairs discrimination task to account for any potential confounding effect of hearing loss on word learning performance. Clearly, if individuals struggled to correctly perceive the phonological label being paired with a novel object, they would be expected to show poor word learning regardless of their verbal short-term memory or phonological awareness skills. This concern was particularly relevant in the context of the current study because hearing difficulties are commonly associated with Down syndrome (see Marcell and Cohen, 1992, Roberts et al., 2007). In the absence of formal audiometric measures of hearing loss from our participants, we used scores on this discrimination task to screen participants for perceptual problems and to compare the two groups for their ability to correctly perceive phonological information.

In addition, measures of individuals’ existing receptive and expressive vocabularies were taken to compare patterns of receptive (referent) and expressive (form) experimental word learning with individuals’ existing receptive and expressive word knowledge. Finally, a version of the Raven’s Coloured Progressive Matrices (CPM) (Raven, Raven, & Court, 1998) was administered to provide a nonverbal index of individuals’ general level of intellectual development.

A sample of 64 typically developing children, within the age range of 5 to 8 years, was assessed to provide a normative picture of the development of new word learning skills and their association with the other measures employed here. A smaller sample of 22 individuals with Down syndrome was also seen, and these individuals’ performance was compared with that of the larger sample of typically developing children (cf. Gupta, MacWhinney, Feldman, & Sacco, 2003). The two groups were loosely selected to be of a comparable level of existing vocabulary knowledge but were not explicitly matched at the outset of the study, thereby avoiding issues of generalizability that are inherent in samples that are specifically selected for their level of performance on any one measure (Jarrold & Brock, 2004). Indeed, the relatively larger sample size of typically developing children allowed us to examine correlations between measures in this group and, therefore, to ask broad-reaching questions about the extent to which measures of verbal short-term memory and phonological awareness relate to individuals’ ability to learn the form and referents of novel phonological items.