• Keine Ergebnisse gefunden

Chapter 6 Focus

6.2 Speech production

6.2.4 Results

This section presents the results of the focus elicitation study. For the statistical analysis of the data I used a generalized linear mixed effect (GLME) model with the fixed factors F-ARGUMENT(subject vs. object) and F-TYPE

(non-identificational vs. corrective) and the random factorsSPEAKER and

ITEM (only intercepts) by using the glmer function from R’s lme4 library (Bates et al. 2015). In the next step, I compared the full model including the interaction of the two factors to a reduced model without the interaction by using likelihood ratio tests of the function anova. If the goodness of fit test revealed a significant effect of the interaction, pairwise post-hoc Tukey tests were conducted in order to investigate the differences between the conditions.

If the interaction appeared to be not significant, further likelihood ratio tests compared the relative fits of the model without the interaction to (i) a model without the factor F-TYPEand (ii) a model without the factor F-ARGUMENT. For each model comparison I report theχ2-score, the degrees of freedom and the p-value, which indicate whether the compared models are statistically different from each other.

6.2.4.1 Turkish

The results of the Turkish participants and the distribution of word orders among the four conditions of the experiment (N/SBJ,N/OBJ,C/SBJ,C/OBJ) are reported in Table 6.2.

Table 6.2: Focus elicitation study: Valid Turkish data NON-IDENTIFICATIONAL CORRECTIVE

SBJ OBJ SBJ OBJ

n % n % n % n %

SOV 11 22.4 43 93.5 41 73.2 43 87.8

OV - - 3 6.5 - - 4 8.2

OSV 38 77.6 - - 11 19.6 2 4.1

OVS - - - - 4 7.1 -

-total 49 100 46 100 56 100 49 100

Table 6.2 illustrates that the Turkish speakers show a preference to realize non-identificational foci (either subject or object) immediately left-adjacent to the verb. Hence, they predominantly produce O[S]FOCV orders if the focus is on the subject, while they primarily produce (S)[O]FOCV orders if the focus is on the object. Consider for instance the examples in (224).

(224) Turkish: Item 11 a. Condition: N/SBJ

Muz-u banana-ACC

[adam]FOC man

yi-yor.

eat-PROG.3

‘THE MANis eating the banana.’ (Tu03)

b. Condition: N/OBJ

Adam man

[muz]FOC banana

yi-yor.

eat-PROG.3

‘The man is eatingA BANANA.’ (Tu04)

Nevertheless, the data in Table 6.2 also reveals some instances of [S]FocOV orders, which indicates that Turkish foci can felicitously occur in other po-sitions than the immediately preverbal one. With regard to the corrective focus conditions, the results show a strong preference for canonical orders ((S)OV) in both conditions. By contrast to non-identificational subject foci, which are in the majority of cases realized in the immediately preverbal slot, corrective subject foci are predominantly realized at the beginning of the sentence. Nevertheless, there are eleven answers in which the subject occurs immediately preverbally. This suggests that corrective subject foci may occur in this position, but do not need to. Compare for instance the examples in (225), which indicate that subject foci may occur either in the beginning of the sentence or immediately preverbally.

(225) Turkish: Item 11 a. Condition: C/SBJ

(Hayir,) no

[erkek]FOC

boy

muz banana

yi-yor.

eat-PROG.3

‘No,THE BOY is eating the banana.’ (Tu09) b. Condition: C/SBJ

(Hayir,) no

muz-u banana-ACC

[bir adam]FOC one man

yi-yor.

eat-PROG.3

‘No,THE MAN is eating the banana.’ (Tu05) Moreover Table 6.2 shows four instances in which the corrective subject focus is realized in postverbal position. However, there is no conceivable explanation for this, since from a grammatical point of view the postverbal position can only host background elements, which means that it should not be possible to realize focused elements in this position (Kılıçaslan 2004: 727).

Therefore, I assume that these four incidences result from perceptual errors.

The impact of the two factorsFOCUSED ARGUMENTandFOCUS TYPE

on the occurrence of OSV orders in Turkish is illustrated in Figure 6.2.

Figure 6.2 shows that the answers of the Turkish speakers reveal a strong difference between non-identificational and corrective subject foci. Whereas non-identificational subject foci are preferably realized immediately prever-bally (O[S]FocV), corrective subject foci are predominantly realized in the beginning of the sentence ([S]FocOV). By contrast, focused objects solely occur left-adjacent to the verb (S[O]FocV), which can be attributed to the fact that this position coincides with the base position of direct objects in Turkish.

SBJ OBJ

0 20 40 60 80 100

%ofOSVorders

non-identificational corrective

Figure 6.2: Focus elicitation study: OSV orders produced by Turkish speakers

The statistical analysis of the data reveals no significant effect of the interaction between the factors F-TYPE and F-ARGUMENT. This result is also confirmed by a model comparison, which shows that the removal of the interaction from the full model does not lead to a significant loss of information (χ2(1)= 2.47,ns). However, the likelihood ratio tests comparing the relative fits of the model without the interaction to a model without the factor F-TYPE2(2)= 30.71,p<.001) and to a model without the factor F-ARGUMENT2(2)= 81.52,p<.001) reveal that both factors are highly relevant in order to explain the deviance of the results. The winning model is presented in Table 6.3. The positive estimates of both factors indicate that OSV orders occur significantly more often with (a) non-identificational foci than corrective foci and with (b) subject foci than object foci.

Table 6.3: Focus elicitation study: Fixed effect estimates for Turk-ish OSV orders

Fixed effects Estimate SE zvalue Pr(>|z|) (Intercept) -6.89 1.23 -5.57 2.43e-08***

F-TYPE 2.67 .57 4.64 3.47e-06***

F-ARGUMENT 5.12 .94 5.39 6.81e-08***

*p< .05;**p< .01;***p< .001

6.2.4.2 Russian

The valid answers of the Russian participants and their distribution among the four conditions of the experiment (N/SBJ, N/OBJ, C/SBJ, C/OBJ) are summarized in Table 6.4.

Table 6.4: Focus elicitation study: Valid Russian data NON-IDENTIFICATIONAL CORRECTIVE

SBJ OBJ SBJ OBJ

n % n % n % n %

SVO 47 75.8 59 100 5 8.1 59 100

OVS 15 24.2 - - 56 90.3 -

-SOV - - - - 1 1.6 -

-total 62 100 59 100 62 100 59 100

The data in Table 6.4 reveal that the Russian speakers show a general preference to realize non-identificational foci in the canonical SVO order.

Consider for instance the examples in (226). Nevertheless, Table 6.4 shows

that there are fifteen instances with the non-identificational subject focus being realized in clause-final position (OV[S]Foc).

(226) Russian, Item 11 a. Condition: N/SBJ

[Muzhchina]FOC man

yest’

eat:IPFV.3.SG

banan.

banana:ACC

‘THE MANis eating the banana.’ (Ru03)

b. Condition: N/OBJ

Muzhchina man

kushayet eat:IPFV.3.SG

[banan]FOC. banana:ACC

‘The man is eatingA BANANA.’ (Ru04)

With regard to the corrective conditions, the participants show a prefer-ence for clause-final foci. They predominantly produce OV[S]Focorders with subject foci and SV[O]Foc orders with object foci. Compare the examples in (227). Furthermore, Table 6.4 reveals five instances of corrective subject foci with [S]FocVO orders as well one verbfinal construction ([S]FocOV), which implies that the position of corrective foci is flexible.

(227) Russian, Item 11 a. Condition: C/SBJ

(Net,) no

banan banana

yest’

eat:3.SG

[muzhchina]FOC. man

‘No,THE MAN is eating the banana.’ (Ru01) b. Condition: C/OBJ

(Net,) no

muzhchina man

yest’

eat:3.SG

[banan]FOC. banana

‘No, the man is eatingA BANANA.’ (Ru06) The amount of OVS orders triggered by the two factorsFOCUSED ARGU

-MENTandFOCUS TYPEis also illustrated in Figure 6.3.

SBJ OBJ 0

20 40 60 80 100

%ofOVSorders

non-identificational corrective

Figure 6.3: Focus elicitation study: OVS orders produced by Russian speakers

The statistical analysis of the Russian data reveals a significant effect of the interaction of the factors FOCUS TYPE and FOCUSED ARGUMENT

(p <.01). This finding is also supported by the model comparison which shows that a model including the interaction fits significantly better to the results than a model without this interaction (χ2(1)= 8.41,p< .05). Consider the winning model in Table 6.5.

Table 6.5: Focus elicitation study: Fixed effect estimates for Rus-sian non-canonical orders

Fixed effects Estimate SE zvalue Pr(>|z|)

(Intercept) -6.08 1.71 -3.55 .00037***

F-TYPE -.02 1.47 -.01 .98506

F-ARGUMENT 10.61 2.63 4.03 5.45e-05***

F-TYPEF-ARGUMENT -6.8 2.58 -2.63 .00851**

*p< .05;**p< .01;***p< .001

Pairwise post-hoc Tukey tests indicate that Russian speakers produced significantly more OVS orders with (a) corrective subject foci than with corrective object foci (p<.001), (b) non-identificational subject foci than with non-identificational object foci (p<.001), and (c) corrective subject foci than with non-identificational subject foci (p<.001), see Table 6.6.

Table 6.6: Focus elicitation study: Pairwise post-hoc comparisons (Russian)

contrast Estimate SE zvalue pvalue

N/OBJ -C/OBJ .027 1.47 .019 1

C/SBJ-C/OBJ -10.61 2.63 -4.035 .0003***

N/SBJ -N/OBJ -3.81 1.19 -3.177 .0081**

N/SBJ -C/SBJ 6.83 2.10 3.239 .0066**

*p< .05;**p< .01;***p< .001

6.2.4.3 Urum

Previous studies on Urum have shown that the word order has undergone a change from OV to a language with a free placement of the V within the VP (cf. Chapter 2). This change becomes also apparent in the descriptions of the Urum participants, consider Table 6.71.

Table 6.7: Focus elicitation study: Valid Urum data NON-IDENTIFICATIONAL CORRECTIVE

SBJ OBJ SBJ OBJ

n % n % n % n %

SVO 33 60 46 83.6 41 70.7 11 57.1

OVS 16 29.1 6 10.9 15 25.9 1 4.8

SOV 3 5.5 3 5.5 - - 7 33.3

OSV 3 5.5 - - 2 3.4 1 4.8

total 55 100 55 100 58 100 19 100

Table 6.7 illustrates that the Urum speakers produced both verbmedial (SVO, OVS) and verbfinal (SOV, OSV) orders. However, the number of verbmedial constructions is significantly higher, which might be explained by priming effects from the context questions (cf. Section 6.2.2.2).

1What is striking is the little number of valid constructions in the corrective object focus condition, which might probably be attributed to the little size of the objects in the context scene and that the participants simply forgot about the objects.

For the further statistical analysis, the V-medial and V-final constructions will be analyzed separately. Consider Tables 6.8 and 6.9.

Table 6.8: Focus elicitation study: Urum V-medial constructions NON-IDENTIFICATIONAL CORRECTIVE

SBJ OBJ SBJ OBJ

n % n % n % n %

SVO 33 67.3 46 88.7 41 74.5 11 91.7

OVS 16 32.7 6 11.3 15 25.5 1 8.3

total 49 100 53 100 56 100 12 100

Table 6.9: Focus elicitation study: Urum V-final constructions NON-IDENTIFICATIONAL CORRECTIVE

SBJ OBJ SBJ OBJ

n % n % n % n %

SOV 3 50 3 100 - - 7 100

OSV 3 50 - - 2 100 -

-total 6 100 3 100 2 100 7 100

Table 6.8 shows that the Urum native speakers have a general preference for SVO orders among all four conditions. Consider for instance the examples in (228). Nevertheless, the data in Table 6.8 indicates that subject foci (both non-identificational and corrective) are more likely to occur with OVS orders than object foci.

(228) Urum, Item 10 a. Condition: N/SBJ

[Ärüf]FOC man

oh-ier

read-IPFV[3]

gazet-i.

newspaper-ACC

‘THE MANis reading the newspaper.’ (Urum08) b. Condition: C/SBJ

Yox, no

[ärgishi]FOC man

yoll-ier send-IPFV[3]

pismo-yi.

letter-ACC

‘No,THE MAN is sending the letter.’ (Urum02)

The impact of the two factorsFOCUSED ARGUMENTandFOCUS TYPE

on the occurrence of OVS orders in Urum is also summarized in Figure 6.4.

SBJ OBJ

0 20 40 60 80 100

%ofnon-canonicalorders

non-identificational corrective

Figure 6.4: Focus elicitation study: OVS orders produced by Urum speakers

The statistical analysis of the Urum data reveals that a model including the two-way interaction of the factors F-TYPE and F-ARGUMENT is not significantly different from a model without this interaction (χ2(1)= .15,ns).

Further model comparisons between a model without the interaction to a model without the factor F-TYPE2(2) = .75,ns) and a model without the factor F-ARGUMENT2(2)= 16.16,p<.01) show that only the factor F-ARGUMENT is relevant in order to explain the results, whereas the fac-tor F-TYPE can be removed from the model without a significant loss of information. Consider the winning model in Table 6.10.

Table 6.10: Focus elicitation study: Fixed effect estimates for Urum OVS orders

Fixed effects Estimate SE zvalue Pr(>|z|) (Intercept) -2.56 .57 -4.43 9.38e-06***

F-ARGUMENT -1.47 .51 2.94 .00323**

*p< .05;**p< .01;***p< .001

The V-final constructions summarized in Table 6.9 reveal that subject foci induce more OSV orders than object foci. Moreover, the data in Table 6.9 show a difference between the two focus types: Whereas non-identificational subject foci induced both [S]FocOV and O[S]FocV orders, corrective subject foci only triggered O[S]FocV orders. Moreover Table 6.9 reveals one instance of an OSV order with a corrective object focus. Consider the example in (229).

(229) Urum, Item 11, Condition: C/OBJ

[Banan]FOC banana

muzhchina man

yest’.

eat-IPFV[3]

‘The man is eatingA BANANA.’ (Urum02)

The example in (223) is of particular interest because it implies that the position of foci in Urum is flexible, i.e., foci may not only occur postverbally (OV[S]Foc) or immediately preverbally (O[S]FocV), but also in the begin-ning of a sentence ([O]FocSV). However, obviously the number of V-final constructions is too small in order to draw reliable results.

In sum, the results of the focus elicitation study provide evidence for the assumption that the position of foci in Urum is depending on the type of constructions that the speakers use: When using V-medial constructions, foci occur either in the beginning of the sentence or clause-finally. When using V-final constructions, foci occur either in the beginning of the sentence or immediately preverbally. The statistical analysis of the V-medial construc-tions shows that only the FOCUSED ARGUMENT(subject vs. object) has a significant effect on the appearance of clause-final foci in Urum, while the factorFOCUS TYPEhas no significant effect. However, the factor might have an effect on the occurrence of immediately preverbal foci. However, this could not be proved due to the little amount of V-final constructions.