In Cognition

Casual speech processes, such as /t/-reduction, make word recognition harder. Additionally, word-recognition is also harder in a second language (L2). Combining these challenges, we investigated whether L2 learners have recourse to knowledge from their native language (L1) when dealing with casual-speech processes in their L2. In three experiments, production and perception of /t/-reduction was investigated. An initial production experiment showed that /t/-reduction occurred in both languages and patterned similarly in proper nouns but differed when /t/ was a verbal inflection. Two perception experiments compared the performance of German learners of Dutch with that of native speakers for nouns and verbs. Mirroring the production patterns, German learners' performance strongly resembled that of native Dutch listeners when the reduced /t/ was part of a word stem, but deviated where /t/ was a verbal inflection. These results suggest that a casual speech process in a second language is problematic for learners when the process is not known from the leaner's native language, similar to what has been observed for phoneme contrasts. Running head: NON-NATIVE PERCEPTION OF /t/-REDUCTION 3 The role of native-language knowledge in the perception of casual speech in a second language Speech perception studies are often performed under ideal circumstances, with participants listening to their native language in a sound-proof booth. Experimental stimuli have usually been recorded under similarly optimal circumstances by a speaker, who is carefully reading out loud. Outside the laboratory, the situation is, however, often less ideal. Environmental noises are common, speakers are less careful than during reading, and the language we listen to might not be our native language. All of these influences make speech perception harder. To start, speech perception in a second language is notoriously difficult. Nonnative listeners have the disadvantage that the second language (L2) often contains phoneme contrasts that they are unfamiliar with given their native language (L1). A large body of research (see, e.g. Strange, 1995 and Bohn and Munro, 2007 for an overview). has shown that mismatches between the phoneme repertoires of the L1 and L2 cause perceptual difficulties. A classical example is the segmental contrast between /r/ and /l/ in English, which difficult for Japanese listeners because these speech sounds match a single Japanese category equally well, leading to perceptual difficulties (e.g., Bradlow, Pisoni, Akahane-Yamada, & Tohkura, 1997; Cutler, Weber, & Otake, 2006; Ingram & Park, 1998; Underbakke, Polka, Gottfried, & Strange, 1988). Cutler and Otake (2004), tested auditory repetition priming with English /r/-/l/ minimal pairs—such as rightRunning head: NON-NATIVE PERCEPTION OF /t/-REDUCTION 4 light—and Japanese listeners. This contrast is difficult for these listeners and consequently priming was observed. These difficulties of L2 listeners are not limited to minimal pairs. Research in psycholinguistics has shown that the recognition of any word entails activation of multiple word candidates which compete for recognition: For example, when listeners hear the word captain, not only the word captain is activated, but also words such as cap and capitol are temporarily activated (Davis, Marslen-Wilson, & Gaskell, 2002). During this process, difficult L2 contrasts create additional competitors for non-native listeners. For instance, Dutch learners of English, who have a hard time to distinguish the vowels /ɛ/ and /æ/, activate the word deaf /dɛf/ when hearing the longer word daffodil /dæfədɪl/ (Broersma & Cutler, 2011). As this shows, listening to a non-native language is, compared to L1 listening, an adverse condition by itself. Non-native listeners are confronted with difficult segmental contrasts leading to spurious lexical activation. To make matters worse, L2 perception seems to be more strongly affected by additional challenges. For example, several studies have shown that, in the presence of background noise, speech perception by non-native listeners is more strongly affected than that of native listeners (see Lecumberri, Cooke, & Cutler, 2010, for an overview). Non-native listeners do not only require a better signal-tonoise ratio (SNR) for the recognition of speech in noise, they also need larger changes in the SNR to recover (i.e., their psychometric functions of recognition rate over noise levels are shallower than those of native listeners, see Van Wijngaarden, Steeneken, &