Phoneme discrimination is an important factor in language acquisition. Within the first year of life, infants turn their early global discrimination skills into language-specific tools, facilitating lexical growth. However, word learning impedes phoneme discrimination between 14 and 17 months of age (e.g., Stager & Werker, 1997; Yoshida, Fennell, Swingley, & Werker, 2009).
Phoneme discrimination can also be demonstrated by studying event-related potentials (ERPs) (e.g., Rivera-Gaxiola, Silva-Pereyra, & Kuhl, 2005). The typical ERP-component that responds to a deviant in a series of standard stimuli, thereby indicating discrimination between standard and deviant, is called mismatch negativity (MMN). MMN is an early ERP-component at around 150 to 250 ms after stimulus onset, representing automatic processing on a pre-attentive level (Näätänen, Paavilainen, Rinne, & Alho, 2007). It can be elicited in a passive listening paradigm and is therefore especially suitable for infant studies. However, even higher-level cognition such as the processing of lexical information can be reflected in the MMN (Shtyrov, Hauk, & Pulvermüller, 2004). In 5-year-old children, lexical processing as part of the MMN was indicated at a latency of 400 to 450 ms after stimulus onset (Korpilahti, Krause, Holopainen, & Lang, 2001).
In the present study, a simple auditory syllable discrimination task is converted into a word discrimination task by supplying semantic content to standard and deviant syllables. The first experimental block contained 50 discrimination trials with four to seven repetitions of the standard syllable with an interstimulus interval of 500 ms. The stimuli consisted of one exemplar per syllable [be] and [de] (400 ms duration each), recorded in infant-directed speech by a female speaker. In a second block (50 naming trials) these syllables gained a semantic dimension by associative pairing with two different jackalope soft toys. Four different pictures per soft toy were presented together with the naming syllable in randomised order. The third experimental block equalled the first. As designed for toddlers, the study was kept to a suitable duration of about 10 min.
It was predicted that the first block elicits a typical MMN, indicating discrimination on a phonemic level, but in the third block, a greater latency in the MMN was expected, indicating discrimination on a lexical level. Adult pilot data (N=6), collected to evaluate the extremely short MMN-paradigm and provide a reference group, revealed a strong frontocentral MMN candidate for discrimination on a phonemic level in the first experimental block with deviant negativity between 180 and 250 ms, as well as on a lexical level in the third block, indicated by a deviant negativity with a latency of 360 ms. This shows also that just 50 trials can elicit a MMN, setting the stage to test toddlers. The results of 20-month-olds are predicted to conform to the adult pilot data, possibly with a greater latency both in the phonemic and the lexical MMN. MMN may therefore offer an alternative method to investigate the development of phoneme discrimination in word learning contexts between the ages of 14 and 20 months.