Context Effects in Spoken Word Recognition of English and German by Native
and Non-native Listeners
Robert Albert Felty
Abstract
Spoken word recognition involves integrating acoustic/auditory information
extracted from the signal with linguistic knowledge, including sentential and
discourse context, as well as the frequency of the words in the signal, and
the similarity of target words to other words in the mental lexicon. Recent
research on visual word recognition has shown that morphology may also affect
lexical access, and that the effects of morphology on lexical access may be
language-specific. This study investigates the effect of morphology on spoken
word recognition using two languages which share many phonological
characteristics but differ in key aspects of morphological structure.
Four separate experiments investigated open-set spoken word recognition in
noise using English and German disyllabic words and nonwords, testing both
native and non-native listeners of each language. Results from native
listeners showed facilitatory effects of lexical status and lexical frequency,
as well as inhibitory effects of neighborhood density, consistent with
previous studies using English CVC stimuli. In addition, the results showed a
processing advantage for monomorphemic words over bimorphemic words,
indicating that morphology also has an influence on spoken word recognition.
The processing advantage of monomorphemes was greater for native listeners of
German than of English, which is taken as evidence that the morphological
structure of the language plays a key role in the influence of morphology on
spoken word recognition. Results from non-native listener experiments were
largely consistent with the native listener results, suggesting that
non-native listeners are sensitive to the same context effects as native
listeners, although the size of the context effects were generally somewhat
smaller for non-native listeners, suggesting that the amount of exposure to a
language can also affect processing.
No current models of spoken word recognition can account for all of the effects
found in this study. Full storage models cannot account for effects of
morphology, while morphological decomposition models cannot account for
neighborhood density effects. Therefore, a revised version of the Neighborhood
Activation Model (Luce & Pisoni, 1998) of spoken word recognition is
proposed which posits that words are stored whole in the lexicon, and that in
addition to orthographic, phonological, semantic, and frequency information,
lexical entries also contain morphological information.