Processing of cognates in Croatian as L1 and German as L2 Maja And̄el Jelena Radanović University of Zagreb Laboratory for Experimental Psychology mandel@ffzg.hr University of Novi Sad jelena.radanovic@uns.ac.rs Laurie Beth Feldman Petar Milin SUNY, The University at Albany University of Novi Sad Haskins Laboratories Eberhard Karls University Tübingen lfeldman@albany.edu pmilin@uni-tuebingen.de 1 Introduction the cognate effect. Specifically, most studies find facilitation in the processing of cognates in L2 (Di- Cognates are defined as words similar in form and jkstra et al., 1999; Lemhöfer and Dijkstra, 2004; meaning across two languages. Similarity in form Van Hell and De Groot, 2008), but results are less may range from full orthographic overlap, as in clear when it comes to the effect of cognates in English film – German Film, to partial overlap, L1. For example, Van Hell and Dijkstra (2002) as in English chapel – German Kapelle. Some and Duyck (2005) reported cognates facilitation in pairs of cognate words developed historically from the dominant language, while Kroll et al. (2002) a common ancestor word, whereas others emerge reported small cognate inhibition in an L1 naming when languages come into contact and loan each task, and Caramazza and Brones (1979) failed to other words. Language users are typically un- find such an effect at all. aware of such diachronic pressures. When acquir- In the present study we sought to examine the ing a second language (L2) they can only perceive influence of cognates on lexical processing in a shared elements between L1 and L2. visual lexical decision task, using L1/L2 language Cognates help explain the nature of lexical pro- pairs that belong to different subgroups of Indo- cessing and the manner in which elements from European languages: Slavic L1 and Germanic L2. the two languages interact. Different measures The aim was to carefully replicate recent find- have been used to explore cognate processing ings from a study by Radanović, Feldman, and and representation, including ERP (Midgley et Milin (2014). Crucially, their study showed quite al., 2011; Peeters et al., 2013; Strijkers et al., a complex pattern of effects that included a three- 2009), latencies in single word (Dijkstra et al., way interaction of language (Serbian L1 vs. En- 2010; Lemhöfer and Dijkstra, 2004), and primed glish L2) by cognate status (cognate vs. noncog- lexical decision (De Groot and Nas, 1991), eye- nate) by word frequency (as a numerical predictor movements (Mulder et al., 2011; Rosselli et al., – covariate). Cognates were processed faster than 2012), and scores on standardized tests (Kelley noncognates in L2, but, surprisingly, significantly and Kohnert, 2012; Pérez et al., 2010). Taken to- slower than noncognates in L1. Furthermore, the gether, empirical findings support the claim that size of the effect was greater when word frequency cognates are processed differently from noncog- was low. nate words. Despite the fact that the aforemen- Because this pattern of effects differs from what tioned experimental measures and techniques di- is typically reported in the literature, we designed verge, the conclusion is similar both in language a replication of the Radanović et al. study and production and in language comprehension (Dijk- followed their method and design, this time using stra et al., 2010, for an overview). Nevertheless, another contrasting pair of languages: Croatian results do differ with respect to a range of details, (L1) and German (L2). including the direction as well as the magnitude of Copyright c by the paper’s authors. Copying permitted for private and academic purposes. In Vito Pirrelli, Claudia Marzi, Marcello Ferro (eds.): Word Structure and Word Usage. Proceedings of the NetWordS Final Conference, Pisa, March 30-April 1, 2015, published at http://ceur-ws.org 182 2 Experiment fect Modeling (LMM), in the R software envi- ronment for statistical computing (R Core Team, Late bilinguals of German (N = 69) – students of 2014), with the lme4 and the lmerTest pack- German with Croatian as their L1, participated in ages (Bates et al., 2014; Kuznetsova et al., 2014). a visual lexical decision experiment. There were The refitted model (after removing residual values two forms of the experiment (in Croatian and in greater than 2.5 of absolute standardized units), re- German), and students were randomly assigned to vealed a significant effects of the control predic- one version. The entire experiment (materials and tors, in the expected direction: facilitation from instruction) was in one language and presentation order of a presentation (β = −.044; SEβ = .007; sequence was randomized for each participant. t = −6.42; P r(> |t|) < .0001), and inhibition In preparation for their study, Radanović et from the word length (β = .211; SEβ = .023; al. (2014) also conducted a normative survey with t = 9.33; P r(> |t|) < .0001). Also, there was 1000 Serbian – English translation equivalents a significant effect of the lexicallity of the previ- ranging from pairs consisting of completely ous word, where stimuli preceded by a word were different words (e.g., priča – story) to the identical recognized faster than those preceded by a pseu- cognates (e.g., drama – drama). They then doword (β = −.077; SEβ = .005; t = −14.36; selected 400 noun pairs covering a wider range P r(> |t|) < .0001). of ortho-phonological similarity between L1/L2 Most interestingly, the model revealed a sig- words, using both subjective similarity ratings nificant three-way interaction between word fre- as well as Levenshtein distance. In the present quency, language and cognate status (β = .053; study we made use of 344 of the previously rated SEβ = .012; t = 4.44; P r(> |t|) < .0001). word pairs, and constructed the same number of The observed interaction is an almost exact repli- pseudowords. All of the selected 344 pairs fitted cation of the three-way interaction reported by nicely for the present purposes of studying Croat- Radanović et al. (2014): cognates are processed ian – German cognates, consistently ranging from faster than noncognates in German (L2), but perfect cognates to orthographically different slower than noncognates in Croatian (L1), and the words. We reused the same noun pairs to allow size of the effect is attenuated for high frequency for strict comparisons of the experimental data. words. This pattern of results is depicted in Figure 1. 2.1 Results With regards to the random-effects structure, by-participant and by-item adjustments to the We calculated normalized Levenshtein distance intercept significantly contributed to the model’s measure for pairs of nouns used in two forms of goodness-of-fit. Word frequency and trial order the present experiment. Similarly to the study of needed additional by-participant adjustments for Radanović et al., the distribution of the Leven- the slopes. Similar by-participant adjustments stein distance measure was strictly bimodal, and, for the slope were held by the word length, as before, the modes matched cognate vs. noncog- which also revealed significant correlation be- nate distinction. That allowed us to further use tween adjustments for the intercept and the slope a dummy-coded variable cognate (TRUE/FALSE), (r = −.72), indicating that slower and more same as in the original study (Radanović et al., careful participants were slowed less as item 2014). length increased. Furthermore, we transformed the measures to ensure a better approximation to a Gaussian dis- tribution. Word frequencies and word length 3 Discussion were log-transformed, while an inverse transfor- mation was applied to response latencies, follow- Radanović, Feldman, and Milin (2014) suggested ing Baayen and Milin (2010). that cognate facilitation in L2 and inhibition in L1 As a last step, we excluded a small number of might be specific to the particular pairing of first the extreme outliers (0.07%) from further analysis and second language and/or to the level of profi- based on the visual inspection of the reaction time ciency in the L2. Results of the present study show distribution. that the particular L1/L2 combination is not criti- The data were analyzed with Linear Mixed Ef- cal in the sense that the same pattern generalized 183 ence and the constellation of cues available in the learning environment. In particular, knowledge in L1 as well as learning history will determine the degree and style of interference that we encounter when learning an L2. This kind of blocking effect is well documented in learning theory (Kamin, 1969). A blocking effect describes failures of learn- ing that arise when a target cue is presented with another cue whose informativity with respect to an outcome has already been established. Arnon and Ramscar (2012) demonstrated in great detail how blocking may influence L2 acquisition when cues from the two languages are competing for the same outcome (a symbolic lexical representation). Cue blocking does not apply directly to cog- nates, however, because typically, cues are iden- tical and, thus, cannot compete and/or block each other. Nonetheless, Arnon and Ramscar’s general observation regarding the way in which learning is structured helps to make sense of the present findings. All that is needed is to extend it to the distinctive properties of cognates whereby learn- ing entails mapping the very same cues (cognate word forms) onto the same outcome. Further insights derive from the highlighting effect (Kruschke, 2009) on the target cues. First, the theory predicts that contextual (ambient) cues are informative about the learning cues, but not about outcomes (Kruschke and Hullinger, 2010). Therefore, temporal and/or contingency aspects of the situation are useful for discriminating be- tween specific contexts of learning. Second, learn- Figure 1: Three-way interaction language by cog- ing cues can be unambiguous or ambiguous for nates by frequency to reaction time latencies in vi- a particular outcome, and the highlighting effect sual lexical decision task. predicts that early ambiguous and late unambigu- ous cues are more informative (Kruschke, 2009). to another sample of participants (studying L2 as Thus, the availability of either L1 or L2 (but not their major) and another L1/L2 combination. The both) provides a context for a given cognate cue fact that target frequency played an important role (actively present in the sensory input). Given high- seems more compatible with an account based on lighting mechanism, with cognate forms are un- proficiency. ambiguous cues we expect facilitation for a lat- However, to find a general explanation and ter learned outcome. Conversely, ambiguous cues testable hypotheses we turn to learning theory. should facilitate an earlier learned outcome as in Arnon and Ramscar (2012), who investigated how an L1 context and, hence, noncognates ought to adult learners acquire an artificial L2, convinc- be faster in L1 but slower in L2. ingly demonstrated that “the way in which learn- In summary, in the case of ambiguous cues ing is structured has a considerable impact on what highlighting is in essence a blocking effect: gets learned” (p. 302). In general, knowledge firstly learned relationships will be favored. This acquisition is codetermined by discrepancies be- outcome is fully consistent with the account by tween expectations based on our previous experi- Arnon and Ramscar (2012). In the case of unam- 184 biguous cues, such as cognate words, competition 2010. How cross-language similarity and task de- between cues does not emerge and the latter mands affect cognate recognition. Journal of Mem- ory and language, 62(3):284–301. learned relationships will show some preference. Previous research on highlighting indicates that [Duyck2005] Wouter Duyck. 2005. Translation and this pattern might be even more pronounced associative priming with cross-lingual pseudoho- when the cues are verbally (i.e., linguistically) mophones: evidence for nonselective phonologi- cal activation in bilinguals. Journal of Experimen- encoded (Kruschke et al., 2005; Kruschke, 2009). tal Psychology: Learning, Memory, and Cognition, This is what present results confirm as well. 31(6):1340. [Kamin1969] Leon J Kamin. 1969. Predictability, sur- prise, attention, and conditioning. In B Campbell Acknowledgments and R Church, editors, Punishment and aversive be- haviour, pages 279–296. Appleton-Century-Crofts. The research for this paper was financially sup- ported by the Short visit grant 4784 received by [Kelley and Kohnert2012] Alaina Kelley and Kathryn NetWordS-09-RNP-089, as well as by the Min- Kohnert. 2012. Is there a cognate advantage for typically developing Spanish-speaking English- istry of Education, Science and Technological language learners? Language, speech, and hearing Development of the Republic of Serbia grants services in schools, 43(2):191–204. ON179006 and ON179033. Furthermore, we wish to thank the Ministry of Science, Education and [Kroll et al.2002] Judith F Kroll, Erica Michael, Natasha Tokowicz, and Robert Dufour. 2002. The Sports of the Republic of Croatia for supporting development of lexical fluency in a second language. this research within the framework of the project Second language research, 18(2):137–171. 130-1300869-0826 “Croatian and German in Con- [Kruschke and Hullinger2010] John K Kruschke and tact – sociocultural aspects and paradigms of com- Richard A Hullinger. 2010. Evolution of attention munication”. in learning. In Nestor Schmajuk, editor, Computa- tional models of conditioning, pages 10–52. Cam- bridge University Press, Cambridge, UK. References [Kruschke et al.2005] John K Kruschke, Emily S Kap- [Arnon and Ramscar2012] Inbal Arnon and Michael penman, and William P Hetrick. 2005. Eye gaze Ramscar. 2012. Granularity and the acquisition of and individual differences consistent with learned grammatical gender: How order-of-acquisition af- attention in associative blocking and highlighting. fects what gets learned. Cognition, 122(3):292–305. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(5):830–845. [Baayen and Milin2010] R.H. Baayen and P. Milin. 2010. Analyzing reaction times. Journal of Psy- [Kruschke2009] John K Kruschke. 2009. Highlight- chological Research, 3(2):12–28. ing: A canonical experiment. Psychology of Learn- ing and Motivation, 51:153–185. [Bates et al.2014] Douglas Bates, Martin Mächler, Ben [Kuznetsova et al.2014] Alexandra Kuznetsova, Per Bolker, and Steve Walker. 2014. Fitting linear Bruun Brockhoff, and Rune Haubo Bojesen Chris- mixed-effects models using lme4. arXiv preprint tensen, 2014. lmerTest: Tests in Linear Mixed arXiv:1406.5823. Effects Models. R package version 2.0-20. [Caramazza and Brones1979] Alfonso Caramazza and [Lemhöfer and Dijkstra2004] Kristin Lemhöfer and Isabel Brones. 1979. Lexical access in bilinguals. Ton Dijkstra. 2004. Recognizing cognates and in- Bulletin of the Psychonomic Society, 13(4):212–214. terlingual homographs: Effects of code similarity in language-specific and generalized lexical decision. [De Groot and Nas1991] Annette MB De Groot and Memory & Cognition, 32(4):533–550. Gerard LJ Nas. 1991. Lexical representation of cognates and noncognates in compound bilinguals. [Midgley et al.2011] Katherine J. Midgley, Phillip J. Journal of memory and language, 30(1):90–123. Holcomb, and Jonathan Grainger. 2011. Effects of cognate status on word comprehension in second [Dijkstra et al.1999] Ton Dijkstra, Jonathan Grainger, language learners: An ERP investigation. Journal and Walter JB Van Heuven. 1999. Recognition of Cognitive Neuroscience, 23(7):1634–1647. of cognates and interlingual homographs: The ne- glected role of phonology. Journal of Memory and [Mulder et al.2011] Kimberley Mulder, M Dijkstra, and language, 41(4):496–518. T Schreuder. 2011. Are eye-fixations in cognate processing dependent on entropy? In The 17th [Dijkstra et al.2010] Ton Dijkstra, Koji Miwa, Bianca Meeting of the European Society for Cognitive Psy- Brummelhuis, Maya Sappelli, and Harald Baayen. chology [ESCOP 2011]. 185 [Peeters et al.2013] David Peeters, Ton Dijkstra, and Jonathan Grainger. 2013. The representation and processing of identical cognates by late bilinguals: RT and ERP effects. Journal of Memory and Lan- guage, 68(4):315–332. [Pérez et al.2010] Anita Méndez Pérez, Elizabeth D. Peña, and Lisa M. Bedore. 2010. Cognates fa- cilitate word recognition in young Spanish-English bilinguals’ test performance. Early childhood ser- vices, 4(1):55. [R Core Team2014] R Core Team, 2014. R: A Lan- guage and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. [Radanović et al.2014] Jelena Radanović, Laurie Beth Feldman, and Petar Milin. 2014. Cognates and fre- quency in Serbian and English. In The 3rd Rijeka Days of Experimental Psychology – REPSI, page 37, Rijeka, Croatia. [Rosselli et al.2012] Mónica Rosselli, Alfredo Ardila, María Beatriz Jurado, and Judy Lee Salvatierra. 2012. Cognate facilitation effect in balanced and non-balanced spanish–english bilinguals using the boston naming test. International Journal of Bilin- gualism, 18(6):649–662. [Strijkers et al.2009] Kristof Strijkers, Albert Costa, and Guillaume Thierry. 2009. Tracking lexical ac- cess in speech production: electrophysiological cor- relates of word frequency and cognate effects. Cere- bral Cortex, 20:912–928. [Van Hell and De Groot2008] Janet G Van Hell and An- nette MB De Groot. 2008. Sentence context modu- lates visual word recognition and translation in bilin- guals. Acta psychologica, 128(3):431–451. [Van Hell and Dijkstra2002] Janet G. Van Hell and Ton Dijkstra. 2002. Foreign language knowledge can in- fluence native language performance in exclusively native contexts. Psychonomic Bulletin & Review, 9(4):780–789. 186