Number of meanings and concreteness: Consequences of ambiguity within and across languages

Abstract

We examined the effects of concreteness and ambiguity on language processing. In Experiment 1, English–Spanish bilinguals translated words with a single translation. Contrary to past findings, we observed no concrete-word advantage in translation latency. In Experiment 2, English–Spanish bilinguals translated words with one and more than one translation. Words with multiple translations were translated more slowly and showed the typical concrete-word advantage. Words with one translation showed a reversal of the typical concrete-word advantage in latency. Further, concrete words were uninfluenced by ambiguity. In Experiment 3, we explored whether the interaction between concreteness and ambiguity was a general property of the language processing system. Supporting this idea, in a monolingual lexical decision task, we found an interaction between concreteness and number of meanings analogous to the interaction in translation. We discuss the common mechanism that may lead to this interaction in both within and cross-language processing.

Acknowledgements

Part of Experiment 1 was completed in partial fulfilment of the requirements for the degree of Master of Science at PSU awarded to the first author. Portions of this research were presented at the Fortieth Annual Meeting of the Psychonomic Society, the Tenth Annual Meeting of the American Psychological Association, and the Forty-Third Annual Meeting of the International Linguistic Association. Experiment 3 was completed in partial fulfilment of the requirements for the degree of Doctor of Philosophy at PSU awarded to the first author.

Part of Experiment 1 was conducted at the Cemanahuac Educational Community. The support and resources of that institution are gratefully acknowledged. We thank Vivian Harvey, Harriet Goff-Guerrero, and Charles Goff for their assistance. We thank Gustavo Perry for his assistance in translating the language history questionnaire into Spanish, and Ana Teresa Pérez-Leroux for translating the task instructions into Spanish. We thank Tamar Degani, Gary Dell, Annette De Groot, Michele Diaz, Susan Dunlap, Peter Gianaros, Tamar Gollan, Erica Michael, Anat Prior, Erik Reichle, Ana Schwartz, Janet Van Hell, and anonymous reviewers for their helpful comments on an earlier version of this manuscript and Emily Barth, Ryan Gilligan, Adam Issan, José ‘Tony’ Matamoros, Rhonda McClain, Israel Roling, Ana Schwartz, and Heather Shrigley for their invaluable research assistance.

This research and the writing of this paper was supported in part by NSF Grants BCS-0111734 and BCS-0418071 and NIH Grant MH62479 to JFK at The Pennsylvania State University, by a sabbatical grant to JFK from the Nijmegen Institute for Cognition and Information, The Netherlands, and from grants to NT from the Leibowitz Fund, The Department of Psychology at The Pennsylvania State University, the grant-in-aid of research program at Sigma Xi, and by a matching fund grant from The Pennsylvania State University Chapter of Sigma Xi.

Notes

¹In some cases, however, the translation asymmetry has not been obtained (e.g., De Groot & Poot, 1997; La Heij, Hooglander, Kerling, & Van der Velden, 1996), or the influence of meaning factors such as concreteness has been found during both L2 to L1 and L1 to (eL2 translation .g., De Groot et al., 1994). Thus, there remains a debate in the literature regarding the component processes of the two directions of translation, and of the architecture of the bilingual memory system.

²The stimulus presentation duration was selected by adding 2.5 times the standard deviation to the mean L1 to L2 translation latency in two past experiments with participants from a similar population. This resulted in two latencies of approximately 2000 ms. The presentation time was therefore set to 2000 ms in the present experiment based on the assumption that this would be ample time for the participants to translate the stimuli, but not so long as to encourage participants to respond more slowly than they would otherwise.

³The means and standard deviations are not provided in the De Groot (1992) or De Groot et al. (1994) papers. Therefore, we calculated power using the results from the present Experiment 2 for concrete and abstract words with more than one translation (a condition in which reliable concreteness effects were observed). We calculated the sample size that would be needed to detect a difference between these conditions using an alpha level of .05. The sample size in Experiment 1 yields a power of nearly .80 (our n = 41; n required for power of .80 = 44). Because past research suggests that abstract words are more likely to have multiple translations than concrete words (Schönpflug, 1997; Tokowicz et al., 2002), we also calculated power using the means for abstract words with more than one translation and concrete words with only one translation. Under these conditions, the statistical power of Experiment 1 exceeds .90.

⁴Data on the number of translations can be analyzed in a number of different ways. For the purposes of the present experiment, words were classified as either ambiguous or unambiguous (i.e., having only one dominant translation or more than a single translation) using the most general criteria, averaging over the two translation directions. Although it may be the case that that ambiguity is direction-specific, the existence of multiple translations in the opposite direction may reflect some property of language that is important. For example, the word glass translates into Spanish as both vidrio and vaso depending on the meaning that is intended. If you are translating glass, you have two options. If you are instead translating vaso, it is possible that, by virtue of your knowledge of English, that vidrio may become activated. Indeed, our previous research (Tokowicz et al., 2002) demonstrates that words with multiple translations are considered less semantically-similar than words that have only a single translation. Therefore, ambiguity may be important regardless of the particular direction of translation being performed. However, we wished to address directly whether our classification system affected our analyses. To this end, we conducted a hierarchical regression analysis in which we predicted performance based on the number of translations in the particular direction of translation being performed. The pattern of results was identical to what was obtained using the analysis of variance with ambiguity coded generally. We report only the details of the analysis of variance because of the greater availability of post-hoc tests for probing higher-order interactions with analysis of variance.

⁵One might well argue that cognates with more than a single translation are not true cognates because only one of the multiple alternatives is likely to share word form with its translation. In many cases, these words represent language change such that a foreign word is borrowed into the language and eventually becomes commonly used. For example, in Spanish, the word for computer is ‘ordenador’, but the word ‘computadora’ is commonly used in some Spanish-speaking countries and is therefore a cognate in its dominant usage. In the present experiment these materials were included for purposes of representativeness but do not provide a critical test of the central hypotheses.

⁶Note that the range of concreteness ratings differs between Experiment 1 and 2. However, what is most critical for the test of the hypotheses is that the concrete words with one and more than one translation have a similar concreteness rating, that the abstract words with one and more than one translation have a similar concreteness rating, and that the concrete and abstract words differ in their concreteness ratings. Because only the pattern of findings is being compared across experiments, it is not necessary for the same range to be employed across experiments.

⁷The data from these participants were excluded on the basis that they may not have paid sufficient attention to the task. This conclusion was drawn due to their relatively low accuracy on the nonwords (86%). In addition, the overall printed word frequency of the real word stimuli was rather high (86 occurrences per million), and even the words low in frequency are fairly familiar (e.g., trauma, malice). For these reasons, the data from these participants were replaced by participants who were tested subsequently and who met the accuracy criteria.