5.1 Experiment 6: Processing of subject-object ambiguities
5.1.2 ERP results
ERPs
Figure 5.1 shows grand-average ERPs for subject- vs. object-initial active verbs. The visual inspection indicated that object-initial verbs elicited a broad, centro-parietal negativity between approximately 400 and 600 ms post onset of the critical verb.
C3 Cz C4
P3 Pz P4
F4 Fz
F3
Figure 5.1. Grand average ERPs for object- vs. subject-initial case ambiguous structures at the position of the verb (onset at the vertical bar) in Experiment 6 (cf. Experiment 1 in Bornkessel, 2002). Negativity is plotted upwards.
The statistical analysis, carried-out for twenty-four successive 50 ms time windows beginning at -200 ms to 1000 ms (t1 – t24), confirmed these observation. There was a significant main effect of ORDER between 450 and 600 ms (450-500 ms: F (1,15) = 18.49, p < .01; 500-550 ms: F (1,15) = 24.81, p < .01; 550-600 ms: F (1,15) = 9.97, p < .01), as well as a significant interaction ORDER x ROI between 450 and 650 ms (450-500 ms: F (5,75) = 3.81, p < .01;
500-550 ms: F (5,75) = 3.43, p < .01; 550-600 ms: F (5,75) = 5.64, p < .01; 600-650 ms: F (5,75) = 8.04, p < .01). All significant effects were due to an enhanced negativity for object-initial in comparison to subject-object-initial structures. Resolution of the interactions revealed that there was no significant N400 effect at left-frontotemporal regions, and only a small effect
(A) Subject-Object (B) Object-Subject
0.4 0.8 -6
6
µV s
between 450 and 550 ms at right-frontotemporal regions (450-500 ms: F (1,15) = 6.45, p <
.05; 500-550 ms: F (1,15) = 4.49, p < .09). In addition, the negativity was most pronounced centro-parietally and lasted until 600 ms at left-posterior regions (450-500 ms: F (1,15) = 19.04, p < .01; 500-550 ms: F (1,15) = 14.14, p < .01; 550-600 ms: F (1,15) = 12.39, p < .01), and even until 650 ms at central-posterior (450-500 ms: F (1,15) = 20.22, p < .01; 500-550 ms: F (1,15) = 21.17, p < .01; 550-600 ms: F (1,15) = 21.19, p < .01; 600-650 ms: F (1,15) = 5.89, p < .05) and right-posterior regions (450-500 ms: F (1,15) = 26.91, p < .01; 500-550 ms:
F (1,15) = 18.89, p < .01; 550-600 ms: F (1,15) = 21.91, p < .01; 600-650 ms: F (1,15) = 6.89, p < .05). A schematic overview of the significant results is shown in Table 5.2.
ROIs time windows in ms
t9 t10 t11 t12 t13 t14 t15 t16 t17 t18 t19 t20 t21 t22 t23 t24 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 999
FroL #
CenA # ** **
FroR * #
PosL ** ** **
CenP ** ** ** *
PosR ** ** ** *
Table 5.2. Overview of the significant main effects of ORDER in each of the 6 ROIs (frontal-left =
.1.3 Interim discussion
s already pointed out above, object-initial active verbs in comparison to subject-initial verbs
FroL; frontal-right = FroR; central-anterior = CenA; central-posterior = CenP; posterior-left = PosL;
posterior-right = PosR) in 24 successive 50ms time windows (t1 – t24) starting at -200 ms till 1000 ms post-onset of the critical verbs (# = marginally significant (< 0.9); * = < .05; ** = < .01). There was no significant effect for any comparison in the time windows t1-t8 (-200 to 200 ms).
5
A
elicited a broad, centro-parietal negativity between 450 and 650 ms post onset of the critical verb. However, the finding of an N400 for object-initial structures is somewhat surprising because previous studies investigating the reanalysis of subject-object ambiguities in complement clauses showed a P600 (Friederici & Mecklinger, 1996; Friederici et al., 2001).2
2 Furthermore, for ambiguous object-initial structures in relative clauses, a P345 has been observed (Mecklinger, Schriefers, Steinhauer, & Friederici, 1995).
Nevertheless, the present N400 effect was interpreted as an index of reanalysis (for supporting evidence from an SAT-study see Chapter 4.2 in Bornkessel, 2002). In a study by Hopf et al.
(1998), the dispreferred resolution of accusative-dative ambiguities in German also gave rise to an N400-like effect. It has been argued that this negativity is the reflection of a lexical reaccess, which is required in order to assign dative case instead of the preferred accusative case. However, this line of argumentation cannot account for the present N400 effect, because both critical conditions involved the assignment of dative case irrespective of the word order differences.
On the basis of the observation that dative-nominative is a basic non-derived word order
.1.4 EEG frequency analysis
s in the previous experiments (1-3, 5) we applied the three frequency-based measures pattern in German (for example with object-experiencer verbs like ‘gefallen’, ‘to be appealing to’), Bornkessel (2002) suggested that the reanalysis to this word order might proceed without operations pertaining to the syntactic structure. Hence, the observed N400 effect would be a reflection of enhanced processing costs due to a reanalysis that does not involve any restructuring operations whatsoever.
5
A
evoked power (EPow), whole power (WPow) and phase locking index (PLI) for the EEG analysis (cf. Chapter 2.5). All measures were determined by Gabor wavelet analyses in frequency bins of 0.5 Hz (time window –334 to 1000 ms plus 50% tapering window).
Analyses were confined to lower frequency bands (< 6 Hz) for the midline electrodes FZ, CZ, and PZ. Because the ERP analysis revealed that the observed N400 effect had a centro-parietal distribution with a clear central maximum, the present time-frequency plots were confined to electrode CZ. The statistical analysis was carried out as in Experiment 2.
Results
igure 5.2 shows that, for object-initial sentences, there was a pronounced increase of evoked lta band activity (2-3 Hz) in comparison to subject-initial sentences. This increase as confined to the time range of the N400 effect of the corresponding ERP analysis.
PLI value difference for PLI. Note that subject-initial sentences were subtracted from object-initial
T upper delta EPow (averaged
equency bins: 2.0-3.0 Hz; time window: 300-600 ms), there was a significant main effect of RDER for electrode CZ (F (1,15) = 5.87, p < .03). Analyses of the measures WPow and PLI F
upper de w
Furthermore, there appeared to be a concurrent increase in upper delta phase locking.
Figure 5.2 Grand average ERPs and Gabor wavelet-based time-frequency difference plots in the delta band (2-3 Hz) for the object-initial sentences in comparison to subject-initial sentences at electrode CZ (N=16).
The colour scale depicts the magnitude of the wavelet coefficient differences for EPow and WPow and the sentences, thereby indicating relative increases in activity for the latter.
he statistical analyses confirmed these observations. For fr
O
revealed that this upper delta EPow increase was due to a significant increase of PLI (F (1,15)
= 7.76, p < .02). However, there was no significant effect for WPow (F < 1). The same pattern was also observable at electrodes FZ (EPow: F (1,15) = 8.54, p < .02; PLI: F (1,15) = 12.44, p
< .005; WPow: F < 1) and PZ (EPow: F (1,15) = 3.98, p < .07; PLI: F (1,15) = 3.95, p < .07;
WPow: F (1,15) = 1.96, p < .19). A comparison of the respective F-values revealed that both the significant upper delta EPow and PLI increase were more pronounced at FZ and CZ, but only marginally significant at PZ (cf. Table 5.3; for more details see Appendix E4).
Upper delta
ORDER EPow WPow PLI
Fz ** **
Cz * *
Pz # #
icant m f OR o
fo p fr
; time 00 m = < . * =
Table 5.3 Signif ain effects o DER f r the electrodes Fz, Cz and Pz with regard to the three measures applied r the up er delta equency band (frequency bins: 2.0-3.0 Hz window 300-6 s). (# 07; < .05; ** = < .01).
In addition to comparison
subject-initial verbs, the visual inspection of Figure 5.3 revealed an increase in evoked eta activity for subject-initial sentences in comparison to object-initial sentences.
increases in activity for the latter.
Indeed, the statistical analyses revealed a significant difference in evoked theta power (averaged frequency bins: 4.0-5.0 Hz; time window: 200-500 ms) for subject-initial verbs in
omparison to object-initial verbs at electrode CZ (F (1,15) = 7.29, p < .02). As for upper elta this EPow increase was due to a significant enhancement of phase locking (F (1,15) =
the observed evoked upper delta increase for object-initial verbs in to
th
Figure 5.3 Grand average ERPs and Gabor wavelet-based time-frequency difference plots in the theta band (4-5 Hz) for the subject-initial sentences in comparison to object-initial sentences at electrode CZ (N=16).
Note that object-initial sentences were subtracted from subject-initial sentences, thereby indicating relative
c d
5.74, p < .04), whereas there was no significant effect with regard to WPow (F < 1).
Furthermore, the same pattern was observable at electrodes FZ (EPow: F (1,15) = 4.00, p <
.07; PLI: F (1,15) = 9.60, p < .01; WPow: F < 1) and PZ (EPow: F (1,15) = 10.55, p < .01;
PLI: F (1,15) = 5.92, p < .03; WPow: F < 1). However, in contrast to upper delta activity, the comparison of the respective F-values revealed that the theta effect for subject-initial
sentences was more pronounced centro-parietally than at frontal electrode sites (cf. Table 5.4 and Appendix E4).
Theta ORDER
EPow WPow PLI
Fz # **
Cz * *
Pz ** *
Table 5.4 Signific fe R th trodes Fz, Cz and Pz with
regard to the three sures applied for the eta fr ncy band (frequency bins:
4.0-5.0 Hz; time wi 200-500 ). (# = < .07; * = 5; ** = < .01).