Our understanding of the relationship between verbal short-term memory as indexed by nonword repetition and word learning must now incorporate myriad factors that were not as apparent 17 years ago when Gathercole and Baddeley (1989) proposed that “the phonological memory skills tapped by nonword repetition play a causal role in vocabulary development” (p. 211). In particular, successful nonword repetition involves more than the phonological loop, word learning happens by degrees and is influenced by many factors other than phonology, and children with specific language impairment (SLI), who have served as test cases by virtue of consistently demonstrating phonological memory deficits, often exhibit other deficits with the potential to negatively impact word learning. Gathercole (2006) still makes the case for temporary phonological storage playing an important role in word learning, but with several caveats. I would like to add two.

First, when assessing the relationship between the ability to repeat nonwords and vocabulary acquisition, it is important to distinguish between studies using extant vocabulary and those teaching new words to children. Our research examining the initial stage of word learning, known as fast mapping, has generally not found a significant relationship between nonword repetition and fast mapping in young children with SLI. Gray (2004) assessed fast mapping comprehension and production in 20 preschoolers with SLI who were age and gender matched to children with typical development (TD). On the nonword repetition task (Dollaghan & Campbell, 1998) the TD group repeated significantly more nonwords than the SLI group, and comprehended significantly more words on the fast mapping task, but receptive vocabulary and expressive phonology predicted fast mapping comprehension. Similarly, Gray (in press) studied fast mapping in 53 young children (3–6 years) diagnosed with SLI and 53 children with TD matched for age and gender to the SLI group. The TD group scored significantly higher than the SLI group on nonword repetition at ages 3, 4, 5, and 6; but only the 5-year-old TD group outperformed the SLI group on fast mapping comprehension and production. Age and nonword repetition predicted fast mapping production for the TD group, but no variable predicted a significant amount of fast mapping variance for the SLI group.

Gray (2004) also assessed word learning beyond the fast mapping stage as children were taught words in a 1 : 1 play context. The TD group learned to comprehend and to produce more words than the SLI group, and did so in fewer trials. As in the previous study, receptive vocabulary predicted the number of words comprehended. Expressive language and a measure of semantic knowledge were the only significant predictors for the number of words children learned to produce. In contrast, nonword repetition was a significant predictor of the number of trials children required to learn to produce a word. Gathercole (2006) states that poor phonological learning ability will not prevent word learning; rather, learning is slowed because children require more exposures to words before they can produce them. Our results illustrate this, and suggest that nonword repetition may index phonological storage efficiency, which affects word learning efficiency, and this is impacted by a number of other variables.

Second, different nonword repetition tasks may yield different outcomes because of the properties of the nonwords, scoring, and task instructions. For example, because children can use stored lexical representations to help produce nonwords, the more “wordlike” the nonwords are, the easier they are to repeat (Dollaghan, Biber, & Campbell, 1999; Gathercole & Baddeley, 1990). There is also evidence that nonword repetition can improve with practice (Edwards & Lahey, 1998; Gray, 2003). Gray found that preschoolers with SLI and age- and gender-matched peers with TD significantly improved their nonword repetition performance when provided two opportunities to repeat the list of nonwords 1 day apart. Further, SLI group scores improved more than TD group scores with a medium effect size for the SLI group and a small effect size for the TD group. It is also interesting to note that scores for the first and second administrations were significantly correlated for the SLI group (R = .72) but not the TD group (R = .40). These findings suggest the need for a standardized, norm-referenced nonword repetition task with good psychometric properties that will allow research findings to converge on the relationship between phonological short-term memory and word learning.

Consistent with the observation of many language researchers, Gathercole (2006) has broadened her view to consider a phonological storage deficit as just one of the core problems that may prevent normal language development. This fits well with research investigating the neurobiology of SLI that reports a number of brain differences in these children (see Webster & Shevell, 2005, for a review) including atypical patterns of brain development (Gauger, Lombardino, & Leonard, 1997), atypical coordination of activation across brain regions during encoding and recognition tasks (Ellis Weismer, Plante, Jones, & Tomblin, 2005), and differences in auditory processing (Uwer, Albrecht, & von Suchodeletz, 2002). Given these findings and the complexity of brain–behavior relationships, it is not surprising that we cannot easily define subtypes of SLI (e.g., Conti-Ramsden, Crutchley, & Botting, 1997) or prescribe differential treatments. However, it remains important to accurately quantify the skills of individual children, thereby paving the way for more precise participant descriptions and more individualized treatment. In this respect, nonword repetition continues to serve as a valuable tool for researchers and clinicians interested in SLI.