Japanese and German are investigated as examples of two languages whose prosodic sys-tems build a contrastive pair. In this section, the functions and forms of pitch and seg-mental length in Japanese and German are described that are necessary to understand the following studies. In this thesis, Japanese refers to standard Tokyo Japanese and Ger-man to standard GerGer-man. Prosodic variations in dialects are not addressed.
1.3.1 Pitch
In Japanese, pitch accents are used primarily for lexical contrasts. The presence/absence of a pitch accent is an inherent property of a word3and does not have any prominence-lending or discourse function (Beckman and Pierrehumbert, 1986). The meaning of a word changes depending on the position of a lexically specified pitch fall (e.g. /hàshi-ga/ = chopsticksNOM, /hashì-ga/ = bridgeNOM, /hashi-ga/ = edgeNOM, the grave accent indicates the position of the pitch fall, if lexically specified). Phonetically, a Japanese pitch accent is realised as a sharp pitch fall from a high level occurring near the end of the accented mora to a low level in the following mora (Gussenhoven, 2004; Vance, 1987), which is not the case for a German falling pitch accent. If the first word in a phrase does not have an accent on the first mora, or if a word is spoken in isolation, then it starts with a low pitch, which then rises to high over subsequent morae, which is known as initial low (ibid.). Note that only 14 % of homophonic word-pairs are really distinguished by pitch accents in Japanese (Shibata and Shibata, 1990).
This small number of homophonic word-pairs distinguished by pitch accents in Japanese poses the question whether whether Japanese L1 speakers store lexical pitch information together with words and use pitch accent for an efficient word recognition.
This is especially questionable, because the Japanese pitch accent is ranged between two syllables (= a polysyllabic phenomenon), so that it takes longer time for its processing in spoken word recognition (Walsh, 1993) (see details of the study in the next paragraph).
Moreover, Japanese pitch accent patterns vary due to various post-lexical factors, such as word position in a phrase and in compounds (e.g. Hirose and Minematsu, 2004) and they vary across dialects
3 Hereby, a word is meant asjiritsugo(= “an independent word”) in Japanese, that contains lexical meaning as opposed tofuzokugo(= “an ancillary word”) that carries a grammatical function, (Ma-suoka and Tabuchi, 1992).
1.3 Languages in focus: Japanese and German 7 Walsh (1993) posed Limiting-Domain Hypothesis claiming that the syllable is a unit of processing in lexical access. According to her hypothesis, tone is effectively used for lexical access, while pitch accent is not, although both are perturbations ofF0and acous-tically identical. This is because the former is a meaningfulF0defined for every syllable, while the latter is defined only once per word (= defined asjiritsugo, see Subsection 1.3.1).
To verify her hypothesis, she conducted a same-different judgement experiment in which Japanese listeners heard pairs of CVCV words or nonwords which were either same or dif-ferent, either in pitch accent or in one of the four segments. Different judgements were significantly slower for pairs varying in pitch accent than for pairs which varied segmen-tally, irrespective of the position of the segmental difference. Thus even a difference in the final vowel (at which time the pitch accent pattern should also be unambiguous) led to significantly faster responses than the pitch accent difference. Further studies also support the view that pitch accent is not effectively used for word recognition. Otake et al. (1993) found no effects of pitch accent in a syllable-detection task with Japanese listeners: The first CV of a word was perceived equally rapidly and accurately irrespective of whether the word had HLL (e.g. monaka) or LHH (e.g. kinori) accent pattern. Also a study in neuroscience supports this view: Tamaoka et al. (2014) investigated whether L1 Japanese listeners necessarily use pitch accent in the processing of accent-contrasted homophonic pairs measuring electroencephalographic potentials. Electrophysiological evidence (i.e., N400) was obtained when a word was semantically incorrect for a given context but not for incorrectly accented homophones. Their finding suggests that pitch accent plays a minor role when understanding Japanese. In the case of Chinese, the N400 was consistently observed when an incorrectly accented word was embedded in a con-text (Li et al., 2008; Zhao et al., 2011) strongly supporting that tonal information is essen-tial for the word recognition in Chinese.
However, Minematsu and Hirose (1995) reported opposite findings: They conducted gating experiments and found that detection of the pitch accent that has pitch fall on the second mora facilitates the word recognition. An early pitch fall in theF0contour for this type of accent makes it possible to identify the accent before the completion of the word recognition process. Accordingly, prosodic information should be utilised to facilitate the access to the mental lexicon by limiting the searching space. Also Cutler and Otake (1999) conducted a two-choice classification task, a gating task and a lexical decision task. In their experiments, words were successfully recognised exploitingF0 and they conclude that accentual information constrains the activation and selection of
candi-8 General introduction dates for spoken-word recognition. However, their experimental method is questionable as they used only few speakers for the stimuli and they found a talker effect.
I regard the controversial results across the previous studies as empirical support for the claim that pitch accent in Japanese plays only a minor role in word recognition, be-cause the studies testing Chinese L1 listeners report more consistent findings that they exploit pitch information for word recognition regardless of different task requirements.
Braun et al. (2014) examined the ability to store lexical tone testing German, French and Japanese L1 listeners in comparison to Mandarin Chinese L1 listeners and showed that Mandarin Chinese controls had the highest sensitivity, followed by the German partici-pants. The French and Japanese participants showed no sensitivity. Tonal information seems to be stored and processed differently by Chinese and Japanese L1 listeners, even though both languages employ pitch lexically. Further, these apparently different roles of lexical pitch in Japanese and Chinese bring to the discussion about the linguistic ty-pology of tone languages and pitch accent languages. Japanese is classified to the former one, while Chinese or Thai to the latter, even though Japanese is sometimes classified as a restricted type of a tone language (Hyman, 2009). The Japanese pitch accent system is characterised as a mixture of various properties in prototypical stress vs. tone systems (Hyman, 2006, 2009; Hyman and Wilson, 1992).
In German, pitch is not used for a lexical distinction. Instead, the locations of met-rically strong syllables are determined in a word and they contribute to a lexical distinc-tion. The metrically strong syllables serve as docking sites to which pitch accents may be associated, for example to post-lexical information such as syntactic or pragmatic infor-mation (e.g. signalling statement vs. question sentence or a double contrast) (e.g. Braun, 2006; Féry, 1993) as well as to paralinguistic information such as attitude and emotion of a speaker (e.g. Chen, 2005; Gibbon, 1998; Liscombe, 2007; Scherer et al., 1984; Wichmann, 2000). Phonetically, a pitch fall (and also a pitch rise) in German is accompanied with a longer vowel duration and higher intensity because it takes place in a metrically strong syllables. In a stressed-timed language such as German or English, metrically strong (=
stressed) and weak (= unstressed) syllables differ in duration, vowel quality, pitch and intensity (Ladd, 1996).
Moreover, the inventory of phonological accent types in German pitch accents (= six basic pitch accents types) (e.g., Grice et al., 1996) is richer than in Japanese (= only one pitch accent type, e.g. Venditti, 2000).
Since a Japanese pitch accent primarily has a lexical function and there is only one type of the accent, its use and variation for a post-lexical or paralinguistic purpose is
lim-1.3 Languages in focus: Japanese and German 9 ited compared to German. For example, Li et al. (2013) showed that Japanese L1 speak-ers expressed different emotional statuses by varying maximum and minimum as well as mean pitch without changing the phonological form of a pitch accent. German L1 speakers may additionally vary pitch accent types to convey such paralinguistic infor-mation (Bänziger and Scherer, 2005; Gibbon, 1998).
1.3.2 Segmental length and rhythm
Regarding another prosodic property under investigation in this thesis, segmental length, Japanese has more lexical restrictions than German. Japanese exhibits lexical vowel and consonant length contrasts (e.g. /kite/ =come, /ki:te/ = listen, /kit:e/ =cut, all verbs in the imperative form, the colon indicates a long segment). German, on the other hand, exhibits only lexical vowel length contrasts to a limited extent. That is, vowel length contrasts in German are accompanied with the vowel quality difference except for /a/ and /a:/ as in Stadt (= city) with a short vowel andStaat (= state) with a long vowel (Wiese, 2000). Consonant length contrasts are not used lexically in German (but in Swiss German, Kraehenmann, 2001 and Kraehenmann and Lahiri, 2008). Double conso-nants in German are used only in the spelling system and indicate the shortness of pre-ceding vowels. True geminates, consonants containing a syllable boundary and poten-tial word boundary occur only in sandhi in sequences likeErbpacht,gut tun,Schiffahrt, viel leisten,hinnehmen(Goblirsch, 1990, 18) orMitteilung.
Speech rhythm, a more global combination of durations, also differs in both lan-guages. Japanese is classified as a mora-timed language (Bloch, 1950; McCawley, 1968), while German as a stress-timed language, although the validity of the typological cate-gories is critically discussed (e.g. Arvaniti, 2009; Warner and Arai, 2001). In Japanese, a vowel (V) or a consonant-vowel (CV) syllable takes up one timing unit (mora) and all morae have approximately the same perceptual duration (Bloch, 1950; McCawley, 1968, but also see Beckman, 1982; Han, 1962 for controversial findings). A Japanese pitch ac-cent does not trigger a longer duration (Beckman, 1982; Homma, 1981). Hence, it does not affect the mora-timing. In German, stressed syllables occur at approximately regular intervals.
All these differences in the use of the same prosodic property in Japanese and German may become hurdles in L2 acquisition. The next section presents how transfer from one’s L1 to an L2 is predicted in the influential models and theories of L2 acquisition.
10 General introduction