The results of the interrater evaluation show that annotators agreed on a total of 110 intonation contours in Portuguese and 52 in German yes–no interrogatives. This represents a rather small dataset to carry out detailed statistical analysis, which is why the main focus will henceforth be on a qualitative analysis. The aim is to contrast clearly distinct tunes in the varieties of children and of their interlocutors.
5.1. Results for the Portuguese Map Task
This section discusses the contours obtained for the three children separately. This is because each child was exposed at home to a different regional variety of Portuguese. For each child, a description is given according to the stress pattern of the nuclear word and partner in the recording session. Moreover, predictions are provided about the possible intonation tunes they could employ in those conditions. It was found that the children make use of the expected tunes, but also that there is a high degree of variability.
A Kruskal–Wallis test was performed using intonational pattern as the dependent variable and speaker as the independent variable. The intonational pattern variable was significantly affected by speaker, (H(2) = 61.7, df = 2, p < 0.0001), confirming that the speakers used a different set of tunes. Further, Kruskal–Wallis tests were used to investigate whether the interaction with a partner had an influence for each child. For that purpose, intonational pattern was the dependent variable and partner was the independent variable. The results showed that only in the productions of ART, the interaction with a different partner significantly affected his choice of tune (H(2) = 5.9, df = 1, p < 0.01). The choice of tune was not significantly affected by the two partners in the case of LAU (H(2) = 0.5, df = 1, p = 0.4) or MAR (H(2) = 2.7, df = 1, p = 0.1). These results show that ART was the speaker whose speech was most consistently affected by his interlocutors. How this difference played out will be discussed in more detail in the qualitative discussion below.
First, the results for ART and LAU will be discussed as they represent the speakers whose regional varieties are most different from German contours. In the following, contours of expected and unexpected tunes summarised in
Table 3 will be discussed. Additionally, the results elicited with the partners in
Table 4 will be discussed.
Speaker ART, whose regional variety was FLBP and from whom a total of 24 yes–no interrogatives were obtained, was recorded interacting with (1) a Portuguese-dominant speaker from Rio de Janeiro and (2) a German-dominant partner with limited proficiency in Portuguese. In this case, 13 utterances were obtained from the interaction with partner (1), as opposed to 11 utterances from the interaction with partner (2). The predictions for (1) are that ART will preferably make use of a tune ending in an L% boundary tone, since this is the boundary tone used in both FLBP (see
Section 2.1), his regional variety, but also in Rio de Janeiro Brazilian Portuguese, the regional variety of his partner. When the nuclear accent is on a word bearing final stress, truncation is expected and the tune would be produced with H%. For (2), the prediction is that either of the tunes known for FLBP could be used, given that in the regional variety of FLBP, yes–no interrogatives can be marked by a tune ending on either an L% or an H% boundary tone.
Taking together all 24 intonation tunes produced by ART and the stress pattern of the nuclear word, I evaluated the percentage of patterns that were realised according to the descriptions of the FLBP variety given in
Section 2.1. As shown in
Table 3, ART used the expected tunes for his variety in 54% of instances, whereas in 46%, there was variation. Regarding the interaction with his peers, the results for (1) fulfil the prediction since ART used L% boundary tones in 76% of instances. The results for (2) show that ART produced only tunes with an H%. In one instance, this could be set in relation to word final stress associated with truncation.
Figure 2 and
Figure 3 show tunes produced on words with different stress patterns. In
Figure 2, the tune is realised on a polysyllabic word with antepenultimate stress, recorded while interacting with partner (1). This intonation pattern is expected as there is a low rising nuclear accent and a low boundary tone, which are commonly produced in both varieties of the speakers. In
Figure 3, ART uses L* H%, recorded while interacting with partner (2). This tune is similar to the German L* L-H%, and L* H% is a tune from another Brazilian variety (NEBP). This contour is rather unexpected as low nuclear accents are not canonical in the first postonic and neither is a rise in the final postonic syllable in his variety. Since the stress is placed on the penultimate syllable, it was unlikely that the low boundary tone could be truncated.
The second participant, LAU, a speaker of the SPBP variety, was recorded interacting with (1) an SEP-speaking partner and (2) a German-dominant partner whose proficiency of Portuguese was lower. A total of 16 yes–no interrogatives were obtained, with 13 from the session with partner (1), and another three from the session with partner (2). The prediction for (1) is that the speaker would mainly produce the contours typical for SPBP, such as L*+H L% and L+H* L%, which both end in an L% boundary tone. In case the nuclear accent is on a word bearing final stress truncation of L% is expected. However, another prediction is that she might also display accommodation, since in the variety of her partner, SEP, yes–no interrogatives end in an LH% boundary tone. The prediction for (2), in which LAU was confronted with a partner with low proficiency in Portuguese, is that she would produce one of the two tunes characteristic of the SPBP variety.
Taking into account the stress pattern of the nuclear word in the 16 intonation tunes uttered by LAU, I evaluated the percentage of patterns that were realised according to the descriptions of the SPBP variety provided in
Section 2.1. The results show that LAU realised expected patterns in 56% and non-expected ones in 44% of instances. All of her patterns ended on an H%, and those realised as expected were stressed on the final syllable.
Figure 4 and
Figure 5 show two yes–no questions while interacting with partner (1), a speaker of SEP. For example,
Figure 4 shows an expected contour as the final word carries final stress and the L% tone could be truncated, which has been observed in Southern Brazilian varieties.
Figure 5 shows the contour for a yes–no interrogative with penultimate stress which is realised with an L* nuclear accent and an H%. However, it is unexpected that the final L% could be truncated in this example, as the stress is on the penultimate syllable. SPBP is known for displaying intonation contours that end on L% when the stressed syllable is non-final. Moreover, while interacting with partner (2), it was noticed that LAU switched to German in four occasions, providing clarification of a word that was presumably difficult for her partner, as in this Portuguese example:
Posso ir para esquerda? ‘Can I go to the left?’, followed by the German
links ‘left’.
Finally, MAR, who is a speaker of the SEP variety, was recorded interacting with (1) a fellow SEP-speaking partner and (2) a German-dominant partner whose father is a native speaker of NEBP (from Pernambuco state). A total of 70 yes–no interrogatives by MAR were obtained (51 utterances in the interaction with the SEP speaker and 19 in the interaction with the German-dominant speaker). The prediction for (1) is that the child would produce the contour H+L* LH% for all phrases. The prediction for (2) is that the child would either produce the H+L* LH% contour typical of SEP, her Portuguese variety, or she would accommodate to her German-dominant partner because she speaks a different variety of Portuguese.
Taking into account the stress pattern of the nuclear word in the 70 intonation tunes uttered by MAR, I evaluated the percentage of patterns that were realised according to the descriptions of the SEP variety. This was performed by taking into consideration the collapsed labels. Note that H+L* and L* were collapsed to L*. This evaluation indicates that MAR used the expected pattern in 99% of instances. These results show that MAR mostly relies on the same tune L* H%. Considering the collapsed tones, and that both SEP and German display H% boundary tones, it is no surprise that this is the tune the child used.
Figure 6 shows an intonation contour of MAR interacting with partner (1). This example shows a phonetic realisation of L* H%, which corresponds to a tune that was employed by all three children, see
Table 4.
Figure 7 and
Figure 8 show two examples of MAR while interacting with partner (1) in German, produced with the L* H-
∧H% and L* L-H% contours. Note that, albeit in different languages, the contours are quite similar, displaying a low nuclear accent with a rising tone that culminates on a high boundary tone that reaches different degrees of an F0 maxima. In particular,
Figure 6 and
Figure 8 show strong resemblance with regard to the excursion up to the H peak.
5.2. Results for the German Map Task
For the German map-task, the three children were recorded interacting with only one partner. It should be recalled that all three were regarded as fully proficient in SG and therefore able to master the task without any linguistic barriers. Although in general, the prediction for the German map task was that children would produce the typical contours for SG, namely L* H-
∧H% or L* L-H%, the transfer of Portuguese prosodic traits into SG intonation was also investigated. A total of 52 information-seeking yes–no interrogatives were analysed.
Table 5 presents a summary of the results for each child, together with the linguistic profile of the partners who interacted with them, the tunes attested, and the percentage of appearance. A Kruskal–Wallis test with
intonational pattern as the dependent variable and
speaker as the independent variable was then performed on the data. The results show that intonational pattern was not significantly affected by speaker (H(2) = 2.5, df = 2,
p = 0.3), showing that the three different speakers used comparably similar intonational patterns.
ART was recorded interacting with FER, a Portuguese-dominant
8 speaker of RJBP. Arguably, ART could easily have resorted to an accommodation strategy since FER is more proficient in Portuguese than in German. However, at no time in their interaction did the boys switch to Portuguese. In the 20 utterances analysed, ART employed the L* H-
∧H% and L* L-H% contours in 30% and 70% of instances, respectively. Thus the most frequent tune is L* L-H%, as shown in
Figure 9.
LAU was recorded interacting with CAR, who exhibited a low proficiency in Portuguese. In this case, eight utterances were analysed, this time revealing that L* H-
∧H% and L* L-H% were used equally. As shown in
Figure 1, LAU was the child who mostly used other types of pragmatic contexts. A closer look at the data suggests that in addition to yes–no interrogatives, she used contours typical of confirmation-seeking questions.
Figure 10 shows a contour produced by LAU that corresponds to a polite request. These phrases show intonation contours similar to those described by
Grice and Baumann (
2002) as
Bestätigung einer Bekannten Tatsache, ‘confirmation of a known fact’, with the pattern H+!H* L-%, and
beruhigende oder höfliche Aufforderung, ‘reassuring or polite request’, with the contour H+L* L-%.
MAR was recorded with MAI, a speaker of the same variety of Portuguese, SEP. The two girls were proficient in both languages and in the session in German they did not switch to Portuguese at any point. In the 21 utterances that were analysed, MAR made use of L* H-∧H% and L* L-H% patterns in 21% and 79% of the time, respectively. The most frequent tune is, thus, the L* L-H%.