• Keine Ergebnisse gefunden

4. Experiment 1

4.6 Introduction LDT

and ρ = 0.758, N = 114, p < 0.001, respectively), especially for the Spanish group. Again, the smaller correlation coefficients for the German group underline the lower importance of familiarity and frequency effects in L2 Spanish gender assignment in comparison to transparency. Moreover, as we have seen, the accuracy of Spanish subjects was affected by item frequency and there was a strong effect of Level on the accuracy of assignments for both L1 groups. Lemhöfer et al. (2010) claimed that

“just passively receiving correct input from the L2 environment is not sufficient for changing incorrect gender representations” (p. 121). The present data suggest that exposure does seem to have at least some beneficial effect. Furthermore, the positive correlation between word familiarity and gender certainty confirms that especially for the Spanish group the more familiar words get with exposure, the more certain learners are of their gender.

In conclusion, because of the effects of noun transparency found for L2 Spanish, the present experiment showed that transfer can also be mediated by L2 characteristics, while so far studies had mostly looked at effects of L1 characteristics (e.g., Sabourin & Stowe, 2008; Sabourin et al., 2006;

Sagarra & Herschensohn, 2011; cf. chapter 3). The present finding is in line with the results of the earlier described study investigating gender congruency effects by Bordag and Pechmann (2007, cf.

section 3.3.2). They found effects of noun ending transparency on RTs and error rates in picture naming (experiment 3) and a GJT (experiment 4). No transparency effects were found by Salamoura and Williams (2007) but, as noted by the authors, the number of transparent items might have been too low to reveal any effects. Regarding my differential findings for L2 Spanish and L2 German, the following explanation seems plausible: “[…] one important factor might be the opaque nature of the gender system in the second language: In absence of reliable form-related cues (i.e., word endings as in Spanish or Italian) for word gender (like in German or Dutch), the learner tends to use L1 gender information […]. By contrast, when easy-to-learn rules govern the assignment of grammatical gender, L2 influences might be overruled, or might not even arise in the first place.” (Lemhöfer, Spalek, &

Schriefers, 2008, pp. 327-328). Furthermore, the fact that gender transfer effects were obtained for L1 German subjects by Lemhöfer et al. (2010) and Lemhöfer et al. (2008) shows that L1 speakers of German are able to experience gender transfer from their native language. Therefore, the results obtained here are not due L1 characteristics but L2 characteristics. Nevertheless, the present result does not preclude that transfer from German to Spanish might occur with more intransparent and infrequent Spanish nouns or under circumstances of greater time pressure or greater task demands, such as having to compute greater agreement distances (cf. section 2.2). This potential trade-off between L2 gender system transparency or simplicity, task demands, and proficiency is addressed in Experiment 2.

I will now discuss the Method and results of the LDT, investigating possible gender transfer effects in a comprehension task.

4.6 Introduction LDT

The present LDT aims to expand on the study by Lemhöfer, Spalek, and Schriefers (2008), who, to my knowledge, were the first to investigate gender interference in a visual word recognition task. As summarized in section 3.3.2, bilingual gender interference effects have usually been investigated in production tasks36, and more specifically, in PNTs. However, just because a bilingual gender interference effect has by now oftentimes been found in language production, this does not

36 Except for a few studies investigating gender transfer effects in sentence processing which are discussed in the introduction of Experiment 2 (cf. section 5.1).

automatically imply that gender interference also occurs in word recognition. As stated by Costa and Santesteban (2004) “[…] the nature of the processes involved in each are different enough to warrant caution in exporting assumptions from one modality to the other without independent motivation.”

(p. 253).

One crucial difference between word production and word comprehension studies is the direction of the activation flow. In language production, the information flow is top-down, while in comprehension it is bottom-up. This means that in word production, the flow starts at the conceptual level with the formulation of the preverbal message and ends with the phonological output. In word comprehension, the direction of activation is reversed. Here, first the orthographic/phonological information is perceived and finally leads to activation of the corresponding concept. This has important consequences for grammatical gender processing and might, more specifically, have different implications for languages with transparent and intransparent gender systems. In languages with a transparent gender system like Spanish, gender information can be directly activated from the orthographic noun ending. In languages with a more intransparent gender system like German, on the other hand, first the meaning (semantic level) has to be activated for the syntactic information to become available at the lemma level. In word production, the gender information has to be actively retrieved, while in word comprehension, this information can be passively activated. Therefore, it is relevant for models of bilingual language production and comprehension whether the direction of the activation flow has consequences for the gender interference effect and whether the same gender interference effect observed in language production can also be observed in comprehension. Furthermore, in language comprehension a cue to the target language is contained in the stimulus itself, while in language production the target language has to be selected by the speaker. This might affect the level of activation of the target and non-target language and has thus an impact on potential interference processes. Once again, Costa and Santesteban (2004) point out that “[…] the issue of the simultaneous activation of the two lexicons of a bilingual might have different answers in each modality […].” (p. 253). As discussed at length in section 3.3.2, bilingual gender interference effects in language production have mostly been studied with PNTs using NP and bare noun naming.

Also in the monolingual domain, gender priming effects have been studied more extensively in the area of language production than in language comprehension. In the few studies investigating gender priming in (reading) comprehension, effects for valid gender primes are not always obtained, while results showing that invalid primes inhibit processing of the subsequent target noun are more consistent (e.g., Gurjanov, Lukatela, Lukatela, Savic, & Turvey, 1985; Jacobsen, 1999; Jakubowicz &

Faussart, 1998), as explained by Lemhöfer, Spalek, and Schriefers (2008). Following the reasoning of Lemhöfer et al. (2008), in the bilingual situation, L2 determiner primes of nouns that are gender-incongruent across the two languages might be comparable to invalid gender primes in a monolingual situation. That way, this “hidden incongruent” (p. 314) condition should inhibit processing of the target. Since, as mentioned before, inhibition effects caused by incongruent primes are reliably found in the monolingual literature, it might be possible to obtain similar results with primes that are gender-incongruent across languages.

In bilingual comprehension, so far, besides Lemhöfer, Spalek, and Schriefers (2008), few studies have investigated gender-primed processing. In two earlier mentioned studies (cf. section 2.1) using auditory comprehension it was shown that contrary to native speakers, L2 learners were not able to use gender primes to speed up processing. In an auditory LDT, Scherag, Demuth, Rösler, Neville, and

4.6 Introduction LDT

Röder (2004) showed that native English speakers could not take advantage of gender-congruent adjective primes in their L2 German relative to gender-incongruent adjective primes. Native German speakers, on the other hand, exhibited facilitation and inhibition effects in response to gender-congruent and ingender-congruent adjective primes. In a similar fashion, Guillelmon and Grosjean (2001) demonstrated that, contrary to monolinguals, English–French late bilinguals were not influenced by gender-congruent and incongruent adjective primes when asked to repeat the noun of an auditorily presented NP. However, as explained in sections 3.2.2 and 3.3.1, the ability to use L2 morphosyntactic information in general and gender information in particular might also depend on L1 characteristics, such as if the L1 has or lacks gender. English lacks grammatical gender, therefore possibly rendering it more difficult for these bilingual subjects to use this type of syntactic information in L2 processing. So even if native English speakers are not able to use L2 gender cues effectively, native speakers of gendered languages might be able to do so. As explained in section 3.3.2, Lemhöfer et al. (2008) were able to show inhibition effects in bilinguals, who were native speakers of a gendered language, in a determiner-primed LDT. Consequently, a primed LDT is an appropriate tool to investigate the following question: Can subjects use the gender information of the prime to distinguish words from nonwords faster or are they influenced by the L1 gender information?

If gender interference occurs, RTs to the primed incongruent conditions should be slower than to the primed congruent condition, relative to the unprimed condition. For the bare noun interference effect, RTs in the congruent condition should in principle also be faster than in the incongruent conditions, relative to the monolingual control group. However, since the bare noun interference effect has so far not been investigated in an LDT, the effect might also be different from picture naming.

Notably, there is an important parallel between the two comprehension experiments reported in this thesis, the present LDT and Experiment 2. In both the LDT and in Experiment 2, gender agreement has to be processed: in the LDT between determiner prime and noun and in Experiment 2 between anaphor and referent.

4.7 Method LDT