Non-Final Focus Accents in The Speech of Advanced Italian Learners of German

This study investigates the prosodic marking of focus in non-native German. Ten proficient learners of German with Italian L1 were recorded reading aloud 40 sentences containing mostly non-final focused constituents embedded in an adequate question context. Non-final focus accents in L2 German are difficult for Italian learners to produce, especially in broad focus contexts with de-accentuation of final verb forms (cf. Paschke/Vogt, in press), because their native language has a strong positional requirement of rightmostness. Given that both German and Italian use pitch accents for information structuring, i. e. to highlight important information, a correct placement of focus accents might, however, be favoured by narrow focus contexts in which prosodic prominence has to be assigned to one specific constituent. In addition to this main hypothesis, the study investigated whether additional clues (such as prosodic highlighting of the relevant constituent in the L2 question, a corresponding syntactic and prosodic structure between L1 and L2) might increase the success rate. The data shows that advanced Italian speakers of German L2 correctly realize non-final focus accents in more than half of the narrow focus contexts, but that their success rate is not significantly higher than in the broad focus condition and is not affected by the additional clues provided.


Introduction
Intonation is crucial to non-native language: it not only determines a foreign accent and thus, comprehensibility and speaker image, but it also affects linguistic meaning, e. g. the distinction between questions and statements.Further, many languages, including the two under investigation in this paper, Italian and German, highlight the constituent under informational focus in a sentence with the help of a pitch accent, that is, informational focus and (phonetic) focus accent fall together.But although both German and Italian obey the principle of "the togetherness of focus and intonational prominences" (Vallduví 1991: 295) they differ in the way they achieve it.
In Italian, the "togetherness of focus and intonational prominences" is tied to a fixed structural position, i. e. the rightmost position in a phrase.That is, in a broad ("all new") focus context (such as when answering questions like What happened?or What's new?), in normal linear word order, Italian speakers place a focus accent on the rightmost word.In narrow focus con-Linguistik online 72, 3/15 ISSN 1615-3014 90 texts 1 a focal word order tends to be realized which thereby allows the element under focus to be invariably in phrase-final position (see below, section 3.3).
In German, on the other hand (the target language under investigation), non-final focus accents can often arise.In broad focus contexts, the phrase-final part of a complex verb (in a main clause) often fails to show pitch movements, and pitch accents are realized instead on non-final arguments (cf.section 3.2).In addition, in narrow focus contexts, no focus-related word order changes occur, and pitch accents are instead shifted backwards, while the postfocal, given constituents are "de-accentuated" (cf.section 3.3).
Against this background, we used a production experiment to test precisely those contexts which are likely to cause problems to Italian L1 learners of German.Answers were elicited in which learners had to realize non-final focus accents in German in violation to the rigid positional constraint of rightmostness they are exposed to in their L1.The goal of the experiment was to see whether non-final focus accents would be mastered better in contexts with narrow focus than in contexts with broad focus.
Following the assumption that "accent goes on the focused word" (Ladd 1996: 161), in narrow focus contexts (with focus on a single word) Italian learners just have to obey the principle of focus-to-accent (applied also in their mother tongue) and assign phonetic prominence to the informational focus.By contrast, in broad focus contexts (with the whole phrase under focus) they have to single out which word of the focused constituent(s) actually bears the pitch accent.In the presence of parts of a complex predicate in phrase-final position, they have to learn a language-specific rule which requires that in this case accent is normally realized on an argument in non-final position.That is, in both cases -non-final accent in a narrow (a) and in a broad (b) focus context -they must switch off the structural rule of rightmostness (present in their L1): in the former case (a), they can at least rely on the cross-linguistically valid principle of focus-to-accent, while in the latter case (b), they have to learn a language specific rule unknown in their L1.
We also tested some specific conditions of non-final narrow focus in order to check whether additional prosodic clues (echo questions) or L1-like syntactic-prosodic structures (right dislocations, prenominal possessives) could help the learners in their task to place a correct pitch accent in the target language.
The next section gives an overview of previous research in L2 focus accent placement, while section 3 takes a look at the literature regarding the prosodic marking of information structure in German and Italian. 4 contains our hypotheses regarding the production experiment and an explanation of the experimental set-up.The results are reported in section 5 and are discussed in section 6.Finally, section 7 summarizes our findings.
Peter Paschke and Barbara Vogt: Non-Final Focus Accents in The Speech of Advanced Italian Learners of German ISSN 1615-3014 91 non-native focus accent placement is mainly concerned with English as a foreign/second language which we here review first, before moving on to German L2 focus accent placements.
Regarding English L2, Grosser (1997) examined Austrian pupils aged 10-12 in their first two years of English as a foreign language.The author found several accent shifts (away from the expected word) although the accent placement in German would have been exactly the same as in English.For example, to the teacher's question "Is anybody ABsent today?"2 some pupils answered "NObody is ABsent" instead of realizing narrow focus with de-accentuation of the final word "absent" like in German "NIEmand ist abwesend".Similarly, to the question "WHAT'S your NAME?" somebody answered "My name is BERND.What's your NAME?" instead of "What's YOUR name?"In these cases, the accent shift to the phrase-final position signals a broad focus instead of the adequate narrow focus.The unacceptable accent shifts to the right, certainly not determined by the learners' L1 German, are interpreted by Grosser "as over-productivity of end focus, which is the unmarked form in English […] and also in German" (ibd.: 224).The same tendency to accent the last word of an utterance, irrespective of its status as given or new or lexical category is observed by Ramirez Verdugo (2002), who investigated Spanish (L1) upper intermediate speakers of English.Gut (2009) and Gut and Pillai (2014), while analysing non-native L2 English produced by speakers of various first languages also note a preference for end-stress, and a reduced tendency to de-accent final given elements in non-native speakers of English with respect to native speakers.In Gut and Pillai (2014: 298), the authors propose that "the lack of prosodic distinction between new and given information constitutes a general feature of L2 speech that might reflect a universal tendency".O'Brien and Gut (2011) observed that even advanced German learners of L2 English in utterances with IP-initial focus sometimes place the nucleus on the last content word. 3o summarize, studies concerned with the acquisition of English L2 observe a general tendency to accent the rightmost element in a phrase, irrespective of its status as given or new which points to a generalization of an unmarked final pitch accent placement in acquisition.It also shows that rules or constraints concerning focus accent placement that go beyond the phonological requirement of rightmostness related to features like information status are neither easily acquired nor easily transferred from L1 to L2, given that the difficulty also regards learners whose L1 has similar rules of accent distribution, such as German L1 learners of English L2.
The studies of Krahmer and Swerts (2007) and Swerts, Krahmer and Avesani (2002) do not deal with L2 acquisition, but they highlight an important difference in pitch accent realization between two typologically different languages like Italian and Dutch.The experiment they conducted is based on a dialogue game with cards showing different geometrical figures in a particular colour (red triangle, blue square, black square etc.).The game consists of a series of turns: at the beginning of the game, the first of the two players issues a command (in his L1) to a second player to select a specific card.After each turn, the first and second player switch roles: in other words, the player who gave the order in the first turn now receives the order in the second turn.The game was conceived so that the participants had to describe the relevant game cards using only an adjective and a noun with the two words displaying a different informative value in each turn: the resulting contexts are: CG (contrast in the first word, second word given), GC (first word given, contrast in the second word) and CC (both contrast), respectively.The results show that the Dutch speakers de-accentuate the adjective or noun in presence of "givenness", meaning when it had been mentioned immediately before, whereas Italian speakers did not, but instead placed a pitch accent on both content words (e. g. triANgolo NEro 'black triangle') irrespective of the context.As a consequence, in a subsequent perception experiment, Italian listeners who heard a certain expression (e. g. triangolo nero) were not able to guess what the preceding context had been while the Dutch test persons were able to do so in their L1 (cf.Krahmer/Swerts 2007; Swerts/ Krahmer/Avesani 2002: 647-651).
An analysis of a task-oriented dialogue between two speakers of the Roman variety of Italian undertaken by Avesani and Vayra (2005) showed the same lack of de-accentuation.
Exploiting the experimental setting used in Krahmer and Swerts (2007) and Avesani et al. (in press) conducted a similar game, focusing on the prosodic marking of information structure in German and Italian, likewise two typologically different languages.The authors examined L1 speakers of both languages as well as Italian speakers of L2 German and German speakers of L2 Italian.The results confirm for Italian L1 that there is a rigid positional rule that requires pitch accents on the rightmost position irrespective of its information status (given or with new/contrastive focus).German L1 speakers, on the other hand, readily de-accent given information: if the word containing given information is in the final position of the noun phrase, it is mostly de-accented (87%); if it is in the initial position of the noun phrase, it may still be de-accented.While German speakers in L2 Italian always (100%) correctly assign an accent to the rightmost content word, Italian speakers of L2 German rarely (17%) de-accent the final given element.From these data, the authors conclude that -in acquisitional terms -there is negative transfer from Italian L1 to German L2: while in Italian, the distribution of pitch accents depends solely on phonological structure irrespective of information status, in German it is governed also by pragmatics (information status), with pragmatic factors outranking phonological constraints.Avesani et al. (in press) thus ascribe the accent placement in German L2 to the fact that Italian L1 learners merely transfer the phonological structural rules of their native language to German L2.
This difficulty experienced by Italian L1 learners of German in de-accenting elements in phrase-final position is also confirmed by Paschke and Vogt (in press).The authors investigated non-native semi-spontaneous speech (a total of 18 min.)produced by upper intermediate Italian university students of German L2 with regard to focus accent placement.In line with expectations, the test subjects (1 male, 5 female) did well with IP-final focus accents, but had many problems with the de-accentuation of IP-final verb forms in broad focus contexts, e. g.: [mit ihrer familie zu verBRINgen] instead of [mit ihrer faMIlie zu verbringen]. 4In nonfinal narrow focus contexts, too, the de-accentuation of "given" information in phrase-final position appeared challenging, but the number of utterances with unambiguous narrow focus was too small to draw any meaningful conclusion.The present study attempts to address this issue.The prosodic marking of focus in German and Italian

Focus
As stated above, we distinguish between (phonetic) focus accent and informational focus, taking the latter as a category of information structure.Focus is a difficult category to define, involving dichotomies such as topic vs. comment, theme vs. rheme, givenness vs. novelty, focus vs. presupposition which are used in a whole range of different research fields, from linguistics to philosophy and cognition science (cf.Hartmann/Winkler 2013).Our aim here is not to analyse the semantic, pragmatic, cognitive properties of focus (but see Büring 2006;Baumann/Riester 2013, andPeters 2006 for more precise definitions of these categories), and in the present context, we use a very broad definition of focus: from a semantic-pragmatic point of view, we take focus simply as the part of the discourse with the highest informational weight which the speaker wishes to highlight to a large extent by acoustic means.

Broad focus contexts
In our investigation we are concerned with the prosodic marking of the informational focus in a foreign language for which specific, controlled contexts have been elicited by adequate wh-question and answer pairs.The elicited answers contain two types of focus: sentences with broad focus and sentences in which a single constituent is under narrow focus.In broad focus contexts, speakers normally highlight a single constituent (or part of a constituent), called by Uhmann (1991) the focus exponent (FE),5 which extends or projects focus over the whole focus domain.
( Thus, both Italian and German allow for focus projection, and the differences regard which constituents can function as FE, and the structural position reserved for the FE (cf.Uhmann 1991: 195-218 for German; Lombardo Vallauri 2010 for Italian).
Typically, in German, predicates are not employed as FE, and focus lies preferably on arguments.Consequently, in German prosody, focus accent is often non-final, given that the nonfinite verb forms of complex predicates in main clauses as well as the finite predicate in subordinate clauses normally appear in phrase final position (compare below, example 2).This is sometimes described as "de-accentuation", meaning that in "all new" sentences, the rightmost element in the phrase normally functions as FE and should be accented, but in German, if the rightmost element is part of the predicate, the nuclear accent is shifted to the preceding argument.
[karl ist gestern nach berLIN gefahren.]'Why are you so sad?' 'Karl went to Berlin yesterday.' Likewise, in Italian too, focused elements are preferably arguments, but they are located -in normal linear word order -at the right edge of the syntactic phrase.Thus, whereas in German the FE is often in non-final position, in Italian this is normally not the case: ( To sum up, in normal linear word order in Italian, broad focus is realized on the last constituent of a phrase, irrespective of whether it is an argument or a predicate.By contrast, in German, the focus accent is normally shifted backwards, if phrase-final nonfinite verb forms of complex predicates are realized.7This is a cross-linguistic difference which is likely to cause problems for Italian L1 speakers when acquiring German.
Peter Paschke and Barbara Vogt: Non-Final Focus Accents in The Speech of Advanced Italian Learners of German ISSN 1615-3014 95

Narrow focus contexts
As already mentioned, in German, constituents under narrow focus are highlighted with pitch accents which are shifted backwards within a sentence without changing normal linear word order (cf.Uhmann 1991;Peters 2006;Grice/Baumann 2007).The given constituents after the focus are "de-accentuated", the focus accent is again realized in non-final position: ( By contrast, in Italian, normal linear word order is preferably manipulated in order to realize both the coincidence of informational focus with (phonetic) focus accent and the positional requirement that focus accents are realized rightmost (cf. Vallduvì 1991;Lombardi Vallauri 2010;Ladd 1996).Thus, normal linear word order is often changed and the constituent under narrow focus is shifted to the phrase-final position designated for the most important information: (7) chi ti ha dato questa catenina?'Who gave you this chain?' me l'ha regalata [ricCARdo.]to me it has given ricCARdo.'It was given to me by ricCARdo.' In this example, the marked or "focal word order" (Face/D'Imperio 2005) signals that the focused element is under narrow focus and that this is not a broad focus context.At the same time, the change in normal linear word order guarantees the "togetherness of informational and prosodic focus" in rightmost position.
Rightmost focus is the default case in Italian, but in narrow focus contexts, non-final focus accents can also occur (at least in surface structure): One possibility is to dislocate a discourse-given constituent and place it after the focus accent.In literature on Italian intonation, the question of how to analyse these postfocal elements is still under discussion.One possibility is to analyse the right-dislocated, given elements as extrametrical.In this way, analyses which rely on a prosodic template for Italian based on rightmostness can be maintained: the main prominence is still in phrase-final position, and the elements after the focus, outside the prosodic template, are assumed to be without phraselevel metrical or intonational prominence.This claim has been challenged recently by Bocci andAvesani (2011a, 2011b) and Bocci (2013), who present experimental evidence to argue that while post-focal elements in Italian may not show visible pitch variation, they are indeed assigned a phrase level accent; the authors propose to account for the flat and low pitch contour by assuming a low tone (*L).
To sum up, in narrow focus contexts, phonetic focus in German is preferably assigned to a non-final constituent preserving normal linear word order.By contrast, in Italian (in surface structure) non-final focus accents can occur but only in combination with right-dislocation.

Hypotheses and materials
The fact that even proficient learners of German with L1 Italian have difficulty placing focus accents properly has been described by Paschke and Vogt (in press) with regard to semispontaneous speech.Given that there are limited data in the analysed corpus regarding clear cases of narrow focus, a production experiment was carried out on read speech with ten Italian L1 learners of German.In the experiment, two specific types of sentence were tested, in which German non-final focus accents arise: one in which non-final accent (de-accentuation) is governed by pragmatics (by givenness), and one in which "de-accentuation" of phrase-final elements is governed by argument structure ("deaccentuation" of phrase-final parts of a complex verb).If a significant difference emerges between these two types (with more correct accent placements in the case of de-accentuation of given elements), this can be interpreted as evidence that the above-mentioned principle of focus-to-accent9 is of help in the acquisition process, reflecting "some universal (and possibly pre-linguistic) intonational highlighting function" (Ladd 1996: 165, 167).We also wanted to test whether additional pragmatic clues (echo questions) or imitations of syntactic-prosodic L1-structures with non-final focus accent could help the L2 performance.

Hypotheses A. 1, A. 2
We assume that Italian L1 learners of German can identify the informational focus of a sentence on the basis of its semantic-pragmatic content also in the foreign language.Accordingly, learners should also be able to correctly assign the phonetic focus accent to an argument in narrow focus, given that there is only one constituent to highlight and given that in both languages, informational focus is highlighted acoustically by assigning a pitch accent.
In broad focus contexts, however, complications arise for learners because of the need to identify the correct focus exponent (FE).As explained in 3.2, in German, the non-finite parts of the predicate are often placed phrase-final, but do not qualify as FE and therefore are not accented, despite being in the informational focus.In these cases, the semantic-pragmatic content of the sentence alone is not enough to identify the constituent to which phonetic focus is to be assigned.The hypothesis is that in these broad focus contexts, learners will have more difficulty assigning prosodic focus to a non-final constituent compared to narrow focus contexts in which the semantic-pragmatic content of the sentence can guide them.
First of all, we tested non-final broad vs. non-final narrow focus within the verbal phrase (5 question-answer pairs for each condition).Subjects had to answer both, to a narrow focus context elicited through a wh-question and to a broad focus context elicited with the help of an "all-new" question using the same sentence in which only the prosodic contour changed because of the information-structural difference.
If significantly more inadequate focus accent placements occurred in (9 a), this would point to a pragmatic grounding of phonetic focus in L2 speech, possible in narrow focus contexts in which no language-specific structural rules of accent distribution need to be acquired.In fact, whereas in (9 b), the semantic-pragmatic meaning of the sentence directs the focus accent immediately to the focused constituent, this is not the case in (a) (hypothesis A. 1).Secondly, we also tested a narrow focus context in echo questions 10 (5 question-answer pairs; cf. 9 c), which were presented immediately after the corresponding broad focus question (9 a).Since -unlike condition (9 b) -this question type does not contain a potentially misleading accent on the final verb form, but highlights the initial wh-question-word, thus adding the prosodic clue that the focus is on the argument, even fewer inadequate focus accent placements should occur in this condition compared to condition (9 b) (hypothesis A. 2).

Hypothesis B
As explained in section 3.3, non-final focus in Italian is possible in the case of word order alternations, e. g. the right dislocation of background information.Thus, we designed some 10 Questions of this type are designed to check comprehension and are classified in German as "Rückfrage" (Engel 1991: 55) or "Nachfrage" (Zifonun et al. 1997: 115).
Linguistik online 72, 3/15 ISSN 1615-3014 98 German sentences modelling them after the structure of an Italian sentence with non-final focus (with a right-dislocated, discourse-given element), in order to test whether non-final focus is easier to achieve if a prosodic template (matching of syntactic structure and prosodic contour in both languages) from L1 is available (cf. 10 b) compared to a narrow focus context (cf. 10 a) in which learners are guided "only" by the semantic-pragmatic content of the sentence (5 question-answer pairs for each condition). 11Ladd (1996: 179) states that "Adesso faccio scorrere il TUO bagnetto" is "considerably less acceptable" than a construction with right dislocation like "Adesso faccio scorrere il TUO, di bagnetto" ('Now I'll run YOUR bath'), but he does not exclude it.Even Cardinaletti (1998: 19) Again, if there is a better performance in condition (11 c) compared to condition (11 b), it should be interpreted as evidence for positive transfer to L2.If no clear difference between the two conditions is observable, this could be read either an impossibility to transfer marked prosodic contours or as a general tendency to final accents in L2 acquisition.

Subjects
The subjects, 10 Italian L1 learners of German (aged 21-24, 1 male and 9 female), were tested individually in February 2014.All students were enrolled in a BA degree course in foreign languages (German plus another language) at the University of Venice and had recently passed their second year German language exam.All students declared that Italian was their only mother tongue.They also specified that they had been learning German from 5 to 8 years at school and that their actual level within the Common European Framework of Reference for Languages was level B2 (which is also the level required at the end of the second year).
Only one student had studied in Germany for a long period (6 months on an Erasmus scholarship), while two other students reported stays in Germany of 2 and 3 months respectively.None of the other subjects had ever spent more than a month in a German-speaking country.

Procedure
The 40 question-answer-pairs were randomized, 12 but then presented in the same sequence to all test persons with the help of a PowerPoint-presentation and a monitor.The test persons received on-screen-instruction regarding the task including 3 examples.Each test took approximately 10-15 minutes, including the introduction of the task and the test run.The questions had been recorded by the first author of this paper and were reproduced while the written version appeared simultaneously on the computer screen.The answers were presented only in their written form on the screen and the subjects had to read them aloud.The questionanswer pairs were organized into 7 blocks containing 5 or 6 question-answer pairs each.After each set of question-answer pairs, other stimuli were inserted regarding another research question in order to interrupt the test procedure and avoid monotony.All individual test sessions were recorded with an Olympus Digital Voice Recorder DM550 (recording rate: 256 kbps) and saved as MP3 files.

Evaluation
Two evaluation steps were carried out by the authors, both native speakers of German L1 with a professional background in phonetics.In a first step, the raters noted down (independently tua" (ibd.: 44, note 2).In any case, the German sentence could also be interpreted as an imitation of Italian right dislocation as in 'prendo la TUA, di macchina'. 12As mentioned earlier, we still maintained the sequence of the corresponding broad focus and echo questions in hypothesis A. 2 (conditions (a) and (c)).
Linguistik online 72, 3/15 ISSN 1615-3014 100 from each other) the syllable on which they perceived the nuclear pitch accent in each of the 400 responses. 13The raters carefully listened repeatedly to each audio file using both, Audacity and Praat (cf.Boersma/Weenink 2013) and also controlling the pitch contour produced by the speakers.However, the main criterion was the auditory perception.Following this evaluation procedure, an inter-rater mismatch was observed in 77 cases out of 400 (19.25%).In a second step, each rater listened to the unclear cases again and the number of mismatches was reduced to 27 (6.75%). 14he relatively high number of inter-rater mismatches might be due to the tendency of the Italian L1 speakers to use some pitch movement to the right of the nuclear pitch accent (indicated in the following with the small capitals): (12) du siehst SCHLIMM aus. was tut dir denn WEH? mir tut der ARM weh15 This is in line with findings of Bocci andAvesani (2011a, 2011b) and Grice et al. (2005) who found that post-focal constituents are not de-accentuated in Italian, but receive a compressed pitch accent.In similar cases, an L1 German rater might perceive a focus accent either on the word "ARM", because it has the most prominent pitch movement, or on the word "weh", because it has a compressed pitch accent16 which makes it significantly more prominent than it would be in an L1 realization.
Another explanation might be the use of different tones in the two languages.While in German, a broad focus accent in assertions is normally realized as H*+L (cf.Peters 2006), in Italian it will typically take the form of H+L* (Gagliardi/Lombardo Vallauri/Tamburini 2012; see also Grice et al. 2005, for the realization of declarative broad focus accents in different varieties of Italian -Naples, Bari, Florence and Palermo -realized always as H+L*).Associating these two bi-tonal pitch movements with the monosyllabic words "arm weh", we will have in any case a high pitch on "arm" and a low pitch on "weh", but in the case of H+L* (Italian) this means a nuclear accent on "weh", while in the case of H*+L (German) the focus accent is on "arm".If, in addition, there is no clear intensity distinction between the two syllables, it can be quite difficult for raters to decide where the focus/nuclear accent is.
This problem, which shares some features with the question of peak alignment perception in L2 reported by Mennen (2007)  If, on the other hand, there are weak syllables between the possible focus accent positions, as in (13 b), one can determine the nuclear accent placement according to an analysis proposed by Ladd (1996: 173-174): so if in "nudeln heute" the pitch falls after the H*-syllable "nu", reaching the baseline on the postnuclear stressed syllable "heu" (cf.Grice/Benzmüller 1998), the (German) focus accent (H*+L) is on "nudeln".On the other hand if the weak syllable "deln" is high and the pitch falls on "heu", this can be interpreted as an (Italian) focus accent (H+L*) on "heute" with the leading tone (H) on the preceding unstressed syllable. 18ventually, the statistical analysis was run only with those responses in which both raters were able to identify the placement of the focus accent and in which their judgments coincided.These unanimous judgments were then converted to the nominal categories R(ight), if the focus accent fell on the expected syllable, and W(rong), if it did not.

Results
Here we report the results of the statistical analysis for each of the hypotheses introduced in section 4.1.We expected learners to be able to single out the arguments under narrow focus and correctly place an acoustic focus accent on them (also when non-final) given that in both languages, informational focus is highlighted by assigning a pitch accent.We predicted more difficulties in broad focus contexts, given that in these contexts, the whole sentence is under focus and learners have to acquire a language-specific rule of de-accentuation.

non-final broad focus accent (1 b) within the verbal phrase
As expected, in the narrow focus context (1 a) the learners performed better than in the broad focus context (1 b), but the difference was not significant (χ 2 = 1,57, df=1).In context (1 a), 16 out of 18 wrong focus accent placements were on the last content word ("präsenTIEren" in the example), while the remaining were on the article or possessive of the focalized constituent ([EIne kette] bekommen; [MEIN projekt] präsentieren).In context (1 b), all 25 wrong placements were on the last content word.The echo-question did not help the test subjects to grasp the semantic-pragmatic focus of the utterance, but seemed instead to puzzle them.Thus, our hypothesis was not confirmed, (χ 2 = 0, df=1).The wrong focus accents in condition (1 c) were all on the last content word (part of the predicate), except for "EIne kette bekommen" (once) and "MEIN projekt präsentieren" (twice).

(8%)
Table 3: Hypothesis B: Non-final narrow focus followed by right-dislocated de-accentuated constituent (2 a) vs. non-final narrow focus in normal linear word order followed by de-accentuated part of the predicate (2 b) Condition (2 a), which was modelled after an Italian non-final focus with right dislocation, did not help the test subjects perform better in focus accent placement as compared to condition (2 b), with its de-accented final part of the predicate (which has no corresponding Italian structure).Since the number of right and wrong focus accent placements was exactly the same, our hypothesis was not confirmed (χ 2 = 0,00, df=1).In condition (2 a), all the wrong focus accents fell on the last content word, except for two cases which both had nuclear accent on an article ("…EIne stunde heute").In condition (2 b), only one placement was on an article ("… EIne stunde dauern"), while all the others were on the final content word (a verb).

Hypothesis C. 1 focus accent placement not classified Sentence pattern
Right Wrong 3 a.welches auto NIMMST du? -ich nehme das auto [von TOM].final focus accent.This was confirmed and the difference was highly significant (χ 2 = 29.06,df=1, p<0.001).The wrong focus accent placement in condition (3 a) was "ich mag die muS-IK von emma" instead of "ich mag die musik von EMma".The wrong focus accent placements in condition (3 b) were all on the final content word (e. g.AUto, in the example).Sentence pattern (3 b) has no equivalent in Italian with a similar word order because there is no prenominal "Saxon genitive" in Italian, whereas possessives as in pattern (c).are placed to the left of the noun in Italian too and can probably have narrow focus accent.We therefore expected more adequate focus accent placements in sentence pattern (3 c) than in pattern (3 b).This was confirmed, but the difference was not statistically significant (χ 2 = 2.28, df=1).

Hypothesis
The wrong focus accent placements in condition (3 c) were all on the last content word (a noun), except one ( "mir geFÄLLT dein plan" instead of "mir gefällt DEIN plan").

Discussion
Before discussing the single hypotheses, some general comments are in order.Our test subjects, Italian upper intermediate learners of German, do not simply follow a positional rule which forces the focus accent to fall on the last content word of the utterance.Considering all conditions with a non-final focus accent (1 a, 1 b, 1 c, 2 a, 2 b, 3 b, 3 c), the average success rate (of correct focus accent placement) was 53.7%: 54.7% for narrow focus and 48.0% for broad focus contexts (cf.fig.1).In the experiment, we tested different types of clues to focus accent assignment with the aim of discovering which one was more likely to help the learners overcome the positional requirement of rightmostness observable as a general tendency in language acquisition.The clues were based on the focus-to-accent claim which is considered to be a general, crosslinguistically (and perhaps universally) valid principle, based not simply on linguistic, but also on cognitive knowledge (hypotheses A. 1, C. 1).Other clues involved prosody, either in the form of prosodic highlighting of the relevant constituent in the L2 question (hypothesis A. 2), or as prosodic template with corresponding syntactic structure and prosodic contour There is some limited evidence supporting our hypothesis A. 1 (i.e. for a different success rate in broad vs. narrow focus contexts containing non-final focus accents).If the information structure in a sentence in the target language, German, requires phonetic focus accent on a single constituent (in narrow focus contexts) in non-final position, this is achieved better (58% of correct placements) than a non-final focus accent in broad focus contexts (48%).In the latter case, all the constituents are in the focus and non-final focus accent is due to a language-specific rule of de-accentuation, which removes accent on phrase-final parts of the predicate.Comparing the two contexts, it appears to be more difficult for learners to acquire this language-specific rule of de-accentuation than to guarantee the "togetherness of informational focus and prosodic prominence" by disregarding the positional constraint.
However, the difference between the success rates in the two contexts was not statistically significant.Although a significant effect might emerge with larger test populations, it would seem in any case to be very limited.The results of our experiment thus confirm those of other studies in which a stereotypic final placement of focus accents has been observed and interpreted as a general feature of L2 acquisition, irrespective of the learner's first language (cf.Grosser 1997;Verdugo 2002;Gut 2009).This might have to do with planning problems in spontaneous speech, but also with an insufficient anticipation of the whole sentence in reading experiments.The fact that even learners with German L1 do not always use de-accentuation of given information in L2 English (cf.Grosser 1997) seems to indicate that semanticpragmatic factors play a limited role, at least at some stages of language acquisition, in focus accent placement.
Furthermore, when learners do overcome the stereotypic phrase-final focus accent, their performance might be governed by both general, cognitive principles like focus-to-accent, and by L2-specific rules like the de-accentuation of final verb forms.This would explain why the rate of incorrect placements is very similar for both de-accentuation of final verb forms in broad focus utterances and de-accentuation of final given elements in narrow focus sentences.
This study thus confirms some informal observations of Paschke and Vogt (in press) who examined semi-spontaneous speech produced by a similar population and found limited but contradicting data indicating learners' ability to highlight non-final narrow focus correctly: they sometimes succeeded, and they sometimes failed.
While there is only limited evidence for the focus-to-accent principle, based on the semanticpragmatic content, the other prosody-based clues tested proved to be of no help in the acquisition task.Hypothesis A. 2 was not confirmed: in other words, the percentage of correct focus accent placement in non-final narrow focus sentences followed by a non-finite part of the predicate does not depend on the question type (whether it be normal wh-question or echoquestion with strong accent on the wh-word used to check understanding of a very specific part of the partner's utterance).This result is quite surprising, because the emphatic accent on the interrogative pronoun ("wie bitte?WAS musst du präsentieren?")might be considered an excellent prosodic clue to the fact that only the corresponding constituent of the answer ("ich muss [mein proJEKT] präsentieren") is focused and should therefore be highlighted acousti- cally.Although the majority of answers (54%) displayed correct focus accent placement, contrary to our expectations, this score was not significantly higher than that of the normal whquestion (58%).
According to our hypothesis B, non-final narrow focus accents should again be easier to achieve if a prosodic-syntactic template from L1 is available compared to narrow focus contexts in which learners are guided "only" by the semantic-pragmatic content of the sentence.Therefore, we compared the success rates in answers with the sentence final position occupied a) by a right dislocated adverb (with a corresponding syntactic-prosodic structure in L1 Italian), vs. b) by a verb form (past participle or infinitive).Contrary to our expectations there was no better performance in condition a), but the success rates were identical in both question-answer patterns (58%), as was the failure rate (34%, almost always due to an accent on the final content word).In other words, a prosodic template available in L1, which is connected to a specific syntactic and informational structure, does not affect the success rate of nonfinal focus accent placement in corresponding L2 utterances.Apparently, such patterns cannot easily be transferred from L1 to L2, probably because they are marked, and marked structures are neither easily acquired nor transferred from L1 to L2 (cf.Eckmann 1987, and section 6).Another explanation might be that, as claimed in recent literature (cf.section 3.3), Italian post-focal constituents are not really de-accentuated, but receive compressed pitch accents.In this case there would be no prosodic template to be transferred from L1 to L2.Although the two conditions failed to show any difference in success rates, they are both characterized by a considerable amount of correct accent placement.The explanation for this must lie in the semantic-pragmatic principle of focus-to-accent, while the incorrect placements can once again be attributed to the negative transfer of the phrase-final L1 default accent, to general principles of language acquisition and to the specific situation of a reading experiment in which the lack of anticipation favors the employment of the unmarked final focus accent.
Hypothesis C. 1 regarded two different conditions of focus accent placement in sentences with narrow focus within the nominal phrase.The assumption that narrow focus in sentencefinal position (success rate 90%) is easier to produce than in non-final position (success rate 44%) proved to be correct, with high statistical significance.Italian L1 learners of German have no difficulty placing focus accents in the target language as long as they are rightmost.This confirms observations reported also by Paschke and Vogt (in press).In our experiment, nearly all the test subjects placed the focus accent correctly in this case, whereas significantly more difficulties arose in non-final position.The result shows that the L2 accent placement in narrow focus contexts is controlled only to a certain extent by the semantic-pragmatic principle of focus-to-accent and that the influence of the unmarked utterance-final accent in L1 Italian is still present.If it were subject only to pragmatics, the success rate should be very high in both contexts.If it were determined instead only by the unmarked utterance-final accent, then the utterance-final placements should be very high in both contexts, whereas they actually amount to 45 (90%) in the case of utterance-final narrow focus and to only 24 (48%) for the non-final narrow focus.The percentage of missing de-accentuation of phrase-final given material within a nominal phrase seems low compared to the very high rate of missing de-  Avesani et al. (in press). 19To sum up, the different distributions of correct and incorrect, final and non-final placements in the two conditions can neither be explained by the influence of semantic-pragmatic factors alone, nor can they be explained without them.
With hypothesis C. 2 we tried once more to prove a positive transfer from L1 Italian regarding non-final focus accent placement within the NP.We claimed that a focus accent on the prenominal Saxon genitive, which has no word order equivalent in Italian, would be more difficult to realize than one on a prenominal possessive.This proved to be true, because the success rate was 56% in the latter context and only 44% in the Saxon genitive condition, but the difference was not statistically significant.Once more, target language sentences moulded in some way on L1 sentences with marked non-final focus accent ("ich nehme [DEIN] auto.")do not seem to have a strong positive influence on focus accent placement compared to sentences without such L1 counterparts ("ich nehme [TOMS] auto.")This speaks once again for the fact that marked prosodic-semantic contours of L1 cannot easily be transferred to another language.

Summary
The basic assumption underlying the present paper was the idea that non-final focus accents in L2 German, which are generally difficult to acquire for Italian learners, could be facilitated by narrow focus contexts.This assumption is plausible because in the case of broad focus, the non-final accent placement requires the command of an L2-specific rule, while for narrow focus it is enough to attain the togetherness of informational focus and phonetic prominence, which seems to be a universal principle or, at least, a common feature of many languages, including German and Italian.However, the evidence in support of this basic assumption was not strong, and could simply be a result of chance.Moreover, even the use of echo-questions, which should give a strong prosodic clue to the focused constituent, did not improve the claimed effect of narrow focus.In our opinion, this does not mean that semantics and pragmatics do not affect the focus accent placement at all (otherwise it would be hard to explain why the success rates in almost all non-final narrow focus conditions amount to 54%-58%), but it does suggest that the pragmatics of narrow focus probably do not represent an essential advantage as compared to broad focus contexts.There may be two reasons for this: firstly, a certain number of responses in all contexts would seem to be governed by a default finalaccent pattern resistant to the semantic-pragmatic principle of focus-to-accent (and perhaps boosted by insufficient anticipation); secondly, when overcoming this obstacle, learners at a certain level of proficiency might have acquired not only the rule that demands togetherness of focus and pitch accent (regardless of final position), but also a rule for establishing the focus exponent in broad focus contexts.A simple inter-language rule to cover both cases might sound like this: "Assign the focus accent to the same word as in L1, even if it is not in the sentence-final position!" narrow focus in normal linear word order: a.wo WARST du gestern?-ich bin [zu HAUse] gewesen.'Where were you yesterday?-I was at home.'Non-final narrow focus followed by post-focal and de-accentuated constituent:b.wo WARST du gestern?-ich war [zu HAUse] gestern.'Where were you yesterday?-I was at home yesterday.'Ifthere is a better performance in condition (10 b) than in condition (10 a), this should be interpreted as evidence for positive transfer from L1 to L2.4.1.3Hypotheses C. 1, C. 2Finally, we tested 3 different conditions (with 5 question-answer pairs each) of focus accent placement in sentences with narrow focus within the nominal phrase.In line with the experiments ofAvesani et al. (2013, in press) it was expected that within the nominal phrase, narrow focus in sentence final position (cf.11 a) would be much easier to produce than in nonfinal position (cf.11 b) (hypothesis C. 1).In this case, a substantial number of correct focus accent placements in condition (11 b) could be interpreted in the sense that learners are (at least in part) guided by the semantic-pragmatic principle of focus-to-accent.du? -ich nehme das auto [von TOM].'Which car are you going to take?-I'll take the car of Tom.' Non-final narrow focus within NP b. welches auto NIMMST du? -ich nehme [TOMS] auto.'Which car are you going to take?-I'll take Tom's car.' c. welches auto NIMMST du? -ich nehme [DEIN] auto.'Which car are you going to take?-I'll take your car.' Secondly, we tried to build a condition (11 c), in which a positive transfer from L1 Italian could facilitate a non-final focus accent placement compared to condition (11 b).While the answer in (11 b) with Saxon genitive (ich nehme [TOMS] auto) has no word order equivalent in Italian (prendo la macchina di [TOM]), a prenominal possessive like in (11 c) is possible in Italian, too, and even a narrow focus accent on the possessive (prendo la TUA macchina), though marked, does not seem impossible. 11 Hypothesis C. 1: sentence final focus accent in the nominal phrase (3 a) vs. non-final focus accent (Saxon genitive) in the nominal phrase (3 b) In line with hypothesis C. 1 we expected sentence pattern (3 a) with final focus accent to be much easier for Italian learners of German to produce than sentence pattern (3 b) with non-Peter Paschke and Barbara Vogt: Non-Final Focus Accents in The Speech of Advanced Italian Learners of German ISSN 1615-3014 103

Peter
Paschke and Barbara Vogt: Non-Final  Focus Accents in The Speech of Advanced Italian Learners of German ISSN 1615-3014 105 between L1 and L2 potentially enabling positive transfer from L1 to L2 (matching of syntactic structure and prosodic contour, hypothesis B, C. 2).
17is particularly serious in the case of adjacent syllables with lexical stress (underlined) like in (13 a): Advanced Italian Learners of German

Success rates in sentences with non-final focus accents: narrow focus, broad focus, average in both focus conditions
Another noteworthy finding is that the success rate was very similar in most of the non-final narrow focus conditions, varying from 54% to 58%; only the correct placements in pattern (3 b) with the prenominal Saxon genitive differed, with 44% (cf.fig.2).
Peter Paschke and Barbara Vogt: Non-Final Focus Accents in The Speech of Advanced Italian Learners of German