Behavior Research Methods, Instruments, & Computers 1997, 29 (3), 425-445 Ratings for Welsh words and their English equivalents WILLIAM J. FEAR University ofWales coUegeofCardifj, Cardiff, Wales The rating of English words and their Welsh equivalents provided the opportunity to compare sub­ jective ratings in two languages as well as the opportunity to compare ratings in a deep and a shallow orthography (English and Welsh, respectively). Four variables-age of acquisition (AOA),familiarity, concreteness, and imageability-were rated. AOAand imageability emerged as the two mostimportant extralingual variables (r = .8 and.73,respectively). Although the patterns of ratings were generally con­ sistent within and between languages, some differences did emerge when these patterns were com­ pared with those from other studies. Using similar instructions to rate familiarity and AOAresulted in a low correlation in English (r = - .5) and a high correlation in Welsh (r = - .84). The mean ratings for familiarity, concreteness, and imageability were higher in Welsh than in English (5.23 vs. 3.35, 5.46vs. 4.41, and 5.29 vs. 4.38,respectively). Both ofthese findings are explained in terms of differences in or­ thographic depth, and it is suggested that Welsh may be a more imageable language than English. The majority ofwork on word recognition and produc­ word cannot always be correctly read aloud according to tion has taken place in the English language with mono­ a set of rules for translating graphemes into phonemes. lingual subjects. In the past decade, however, research on English is such a language, with large numbers ofirreg­ word recognition and production has been proceeding in ular words, homophones, homonyms, and so on. The ma­ languages other than English, as well as with bilinguals. jority ofmodels ofreading, and the dominant dual-route A developing bilingual literature has begun to focus on model especially, are based on the English language. The cognitive and psycholinguistic issues (see, e.g., Boyce, dual-route model conveniently fits the behavior ofa deep Browman, & Goldstein, 1987; Caramazza & Brones, 1979; orthography like English, since it proposes that the pri­ Gerard & Scarborough, 1989). In particular, the work of mary route for word recognition is by direct visual ac­ de Groot and others (see, e.g., de Groot, 1992; de Groot, cess, hence allowing irregular words to be read without Dannenburg, & van Hell, 1994; de Groot & Nas, 1991) regularization errors. The phonological route, in which has begun to develop a strong theoretical foundation for words are read using grapheme-to-phoneme correspon­ bilingual word recognition based on shared conceptual dences (GPCs), is used only for reading novel words, in­ representations of words. Bilingualism potentially pre­ cluding nonwords, since they do not have an existing lex­ sents some ofthe strongest challenges--outside ofmulti­ ical representation. lingualism-to existing theories of lexical processing, A shallow orthography is one in which the spelling-to­ which are based on monolingual research. At the same sound rules are consistent. That is, a word can always be time, it has the potential to present powerful evidence in correctly read aloud according to a set ofrules for GPCs. support ofthese theories. Serbo-Croation, Spanish, and Welsh are examples of A related line of developing research compares word such languages. Shallow orthographies have few, ifany, recognition and production in deep and shallow orthog­ irregular words, homophones, or homonyms, although raphies; such a comparison allows an important variable some ofthese word types will almost always occur due to in word recognition-orthography-to be manipulated specific features ofa language. Research that exploits lan­ (see, e.g., Frost, 1994; Frost, Katz, & Bentin, 1987; Katz guages with shallow orthographies is presenting a grow­ & Feldman, 1983). A deep orthography is one in which ing body ofevidence that challenges the assumptions un­ the spelling-to-sound rules are inconsistent. That is, a derlying the existing dual-route models, and even the models themselves (see, e.g., Bosman & de Groot, 1995; Carello, Lukatela, Peter, & Turvey, 1995; Katz & Feldman, This study was based in part on an MSc (Econ) dissertation submit­ 1981; Lukatela, Gligorijevic, Kostic, & Turvey, 1980). ted in partial fulfillment ofthe requirements for the MSc degree at the University ofWales College ofCardiff. The research was supported by Three major features make the case ofWelsh-English Grants K00429413158 and R0042953438I from the Economic and So­ bilingualism particularly useful for this line ofresearch. cial Research Council. I would like to thank Diane Ellis and C. Barry First, Welsh-English bilinguals speak two languages that for their help and support during the research. Thanks are also extended have opposite orthographic depths: Welsh has a shallow to the two reviewers ofthis paper for their comments and suggestions. orthography and English has a deep orthography. How­ Correspondence should be addressed to W.1. Fear, School ofPsychol­ ogy, University ofWales College ofCardiff, PO Box 901, CardiffCFI ever, both languages share a similar structure and have 3YG, Wales (e-mail: [email protected]). almost all the same sounds (D. Brown, 1976; Price, 1984), 425 Copyright 1997 Psychonomic Society, Inc. 426 FEAR although the GPCs are not always the same between lan­ sembling must take place for these words. Thus, AOA guages. Second, as a language, Welsh may behave dif­ measures may indeed make a valuable contribution to ferently from English as a result of its linguistic envi­ language research. ronment. Historically, Welsh is an older and more literary As yet, no one seems to have developed any bilingual language than English. It is also a less used language, measures for AOA. Subjective frequency seems to have having in the United Kingdom around 500,000 speakers; been the primary variable considered in most experi­ 300,000 people who can speak, read, and write it; and mental investigations into shallow orthographies, at least 200,000 people who use it on a dailybasis. Despite its low those studies that have contributed to theory on the in­ rate of use, Welsh is considered a high-status language in fluence oforthography on word recognition. In contrast, parts ofWales, and there is a Welsh language television there seems to be little evidence that the variables AOA, channel, a Welsh language radio station, and Welsh lan­ imageability, and concreteness have been collected for guage schools. A further example is that road signs are equivalent words in two languages, although de Groot presented in both English and Welsh throughout Wales. (1992) and de Groot et al. (1994) did collect, among other A final point on the Welsh language is that the vast ma­ data, subjective ratings for imageability and familiarity jority of Welsh speakers are also bilingual in English for 458 equivalent Dutch and English words. They also (Aitchison & Carter, 1994). There are few, ifany, mono­ had access to objective frequency ratings for both the glot Welsh speakers. Dutch and the English words from the CELEX database A further point that has received much interest but has (see Burnage, 1990). They found extralingual intercor­ not been highly researched (and is not addressed in this relations of .94 for imageability, .63 for familiarity, and work) is that Welsh has a fixed and rule-governed muta­ .78 for objective log frequency. tion system. Mutations occur in Welsh when the initial A number of questions emerge. First, can the same consonants of Welsh words change under some circum­ subjective ratings ofvariables be collected in both deep stances according to fixed and definite rules, without and shallow orthographies? If they can be collected, will changing the meaning ofthe word. Forexample,p changes they show the same pattern of intercorrelations in both to Ib/, Imh/, Iphl in some circumstances. Thus Ipensill languages? Second, if ratings of equivalent words are (pencil) becomes Ibensil/, Imhensil/, Iphensil/. collected from these two orthographies, what pattern of One ofthe important rated variables recognized in the intra- and extralingual intercorrelations will emerge? Fi­ English monolingual literature is familiarity, or subjec­ nally, as would be expected if AOA is the same variable tive frequency, which is sometimes used in place ofob­ as familiarity, will AOAbe highly correlated with famil­ jective frequency counts, concreteness, imageability, and iarity ifthe same instructions are used to collect each of age of acquisition (AOA) (see Carroll & White, 1973; these two variables? The present study begins to address Gilhooly & Hay, 1977; Gilhooly & Logie, 1980; Quin­ these questions, with the additional feature that the lan­ lan, 1992).The termsfamiliarity andsubjectivefrequency guage with a shallow orthography (Welsh) is spoken by are used interchangeably in this paper, and unless stated bilinguals. as objective frequency, any reference to frequency It is assumed that the patterns ofintralingual intercor­ should be taken to indicate "subjective frequency" or relations within each language will mirror each other "familiarity." and, similarly, that patterns ofextralingual intercorrela­ AOA, the age at which people think they acquired a tions will be consistent across languages. Predictions word, has been reemerging as an important variable in cannot be made about levels of correlation, since these cognitive and psycholinguistic research. Ithas been