SOME EXPERIMENTS ON THE ROMAN AND CYRILLIC ALPHABETSOF SERBO-CROATIAN*

Ge Lukatela+ and Me Te Turvey++

COMPARISONS: SERBO-CROATIAN ORTHOGRAPHIES-------AND THEIR SPECIAL PROPERTIES

Much if not most of current theori on the reading process and visualinformation processing is based on investigations with English languagematerialsa such processes vary but little across languages andorthographies and therefore a theory based on one language will suffice foraIle However, what variations there are may prove to be revealinge We havebeen asking whether or not the reading of Serbo-Croatian may make use ofdifferent characteristics of the written word or different encoding routinesthan are used in the reading of Englisha

A distinction that is often made between logographic writing systems,such as Korean, Chinese, and kanj i, and alphabetic systems, such asEnglish and Serbo-Croatian, is that the former refer to the morphology, whilethe latter refer to the phonologyo The logographic system is said to specifyunits of meaning, whereas the alphabetic system is said to specify the soundsof the spoken language, al though the distinction is not as sharpe Indeed,this interpretation of the al is less than ideal as far as English isconcerned, for the between wri tten and spoken English isopaque can be made silent context and, in general, graphemestake on different s in different graphemic contexts. Lookingfor in Gibson, Pick, Osser, and Hammond(1962) advanced the idea , a cluster of letters thatcorresponds to a sound G individual letters in English do not haveinvariant phonemic inter , certain arrangements of letters do, par-ticularly when their locations wi thin words are taken into consideration.Whether or not the notion of is valid, the point is obvious:the cipher relating to utterance in English is complex We argue thatthe cipher in Serbo-Croatian is considerably more transparent; and that forthe Serbo-Croatian the claim that it specifies the sounds ofspeech is to the mark. But let us pursue the Englishorthography a little

NICHD GrantNICHD Grant HD-01

James F G KavPress, 1 pp

LABORATOR ) ]

203

The opaqueness of the script to utterance relation in English is owing,by and large, to two reasons. First, the pronunciation of the languageevolved along different lines from the spelling of the language. Consider thefollowing example ci ted by Henderson (1977). The Engl ish digraph gh as inbough and rough specified a unique guttural utterance until the seventeenthcentury. After the seventeenth century the pronunciation of gh took twodirections: it either became silent as in night or took the phonemicinterpretation IfI as in enough. But the spelling had already becomestandardized largely owing to the efforts of the fifteenth century Englishprinters, such as Caxton; and, in consequence, gh is handed down to thecontemporary reader of English as an orthographic anomaly.

The second reason for the spelling-sound opaqueness is that the Englishorthography may be as close to the morphology as it is to the phonology.Indeed, in the evolution of the English language, Henderson (1977) has statedthat the tendency has been for the orthography to reflect etymology, which istantamount to saying that it reflects the basic units of meaning. In thisvein Chomsky (1970) has argued that the English orthography is near optimalfor wri ting the English language. The orthography preserves the morphology,which would not be the case if the optimality principles were phonemiccorrespondences. Thus, the spelling preserves the following morphologicalsimilari ties--tele-graphy, tele-graph-ic, tele-graphy-y--in the face of theobvious phonetic variability. Similarly anxious and anxiety by virtue oftheir visual likeness permit the reader, in principle, to go directly from theappearance of the letter sequence to its meaning. Therefore, the fundamentalpoint made by Chomsky (1970) and also by Venezky (1970) (but for somewhatdifferent reasons) should be noted, namel y, that the Engl ish orthography issystematic in its own right. It is specific to linguistic structure at a deeplevel and is not to be understood just as a phonemic transcription. Indeed,on the Chomsky-Venezky v the script-utterance relation is opaque selybecause the script and utterance are alternative specifications of the sameunderl ying structure (cf. Francis, 1970) However, the tempering concl usionof Gleitman and Rozin' s (1977) thorough analysis is that it is not so muchthat English orthography is optimal for this or that grain-size of linguisticanalysis, but rather that English writing is a rich mixture of number ofgrains of linguistic representation, together with more than a sprinkl ofarbitrary features

Let us now turn to Serbo-Croatian, Yugoslav ia' s maj or language. Serbo­Croatian, unlike English, is pronounced as it is written; that is, individualletters have phonemic inter that remain consistentchanges in the context in which they are imbedded All written letters arepronounced; hence, in Serbo-Croatian there are no silent and no doubleletters.

This state of affairutterance--is virtue of historicalevolution of the Serbo-Croatian withphy. The modern Serbo-Croatianof the nineteenth century Karadyou and readconstraints on sound quencessequences This contrasts

204

sequences derive not only from phonological constraints but also from a desireto preserve the etymology and graphemic conventions. That is, from a".o.1400-year accumulation of scribal , printing conventions, lexico­graphers' selections, and occasional accident which somehow became codified aspart of the present orthographic (Venezky & Massaro, 1979, p. 25). InEnglish, sequences (such as /wh/) can be orthographically

spell s (such as wh) but no such iari ty is permitted in Serbo-Croatian. -

Karad (1814) selected the in mid-Yugoslavia as the idealand to each phonemic of a letter character or,in a few cases, a combination Karad took the maj ori ty ofletters from the al ex at but since the number of lettersavailable was less than the nmnber of needed, he borrowed or modifiedseveral letters from other al In fact, two alphabets wereconstructed: a Roman alphabet and a Cyrillic alphabet. In modern Yugoslavia,Eastern Serbo-Croatian uses primarily the illic script whereas WesternSerbo-Croatian uses primarily the Roman. In some regions (e.g., Bosnia,Herzegovinia), however, both are used about equallyo

The Serbo-Croatian language In the Cyrillic alphabetthere is one letter for each ; in Roman, 27 phonemes arerepresented by letters and three phonemes by pairs of letters: LJ, NJ,D~. Figure 1 compares the Roman and ic alphabets in uppercase and inTable 1 the letters (both uppercase and cursive) of the alphabets are giventheir correspond letter-names in the International Phonetic Alphabet (IPA)transcription "

An about the Roman and lic al is that they maponto the same but still compri two sets of letters that are,wi th certain , mutuall y excl usi ve. Of the total set 0 f letter scompr the two al the un to one or the otheralphabet (see 1) A number 0 letters, however, are shared by the twoal Of these shared letters, some receive the same phonemic interpre-tation whether read as Roman or lic (referred to as common letters) andsome receive two ic inter , one in the Roman reading and one inthe Cyrillic read (referred to letters) Therefore, one may

ze instances in which letters are different in but pronounced thesame way, e.g., the ic M and the Roman I are both pronounced like the eain seat· instances in which are the same in and pronunciation;

in which of the same butdifferently, e.g., like the n in wine, the RomanH 1

alThree unusualness of Serbo-Croatian bi­

mother, translated into Serbo-Croatianrendered as· [to je aka].

i written in RomanThe sentence

205

Cy r i III C

Serbo- roatian Iphab-Up ercase-

"Commonletters"

Roman

Uniquely

Cyrillic letters

~

Ambiguous

letters

iquely

man letters

206

1. The two al the Serbo-Croatian

Table 1$ Letters of the Serbo-Croatian alphabet.

Serbo~Croa t i ~n

Roman Cyr IlIle

letterPrinted CurSive Printed CurSive Name

Upper Case A A se lower Case In IPA

A (JJ A a aB ~ 6 d beC (" U ~ tseC ; '-i 11 tJaC ; h ;( tlia0 / .n. ,? daf) ,/ l) I d3jaoi A U f d3 aE e E e £

f / cp~ fa

G / r 2 gaH " X oX xaI I l-1 41 i

J J jaK ~ K ;{ kaL ,/ Jl ~ Ie

LJ Jb A> Ije

M m M .M rnaN '" H ~ neNJ til H nja0 £J 0 r- ::>

p / n Ii paR , P / raS d C saS J W 1JIt JaT tf T iii taU N Y uV ~ B va

It .j Z9

i J >+< Ja

207

reading, which is written in Cyrillic l)AK YQl1 ,UA tIHTA but in Roman as, BAKUCI DA ~ITA. Regardless of which alphabet has been used, the phonetictranscription is the same in both cases: [dzjak uci da citaJ, as is themeaning.

A most central feature is that both alphabets are taught in the schoolsand by most accounts the letter forms and the letter-to-sound correspondencesof both alphabets are learned by the end of the second grade. The childrenare taught one alphabet in the first year and a half and then master the otherby the end of the second year. In the western part of the nation the Romanalphabet is learned first and in the eastern part of the nation it is theCyrillic alphabet that the children master ini tially. This geographicallybased ordering of acquisition of the two alphabets provides a model forexamining the relation of two separate symbol systems, learned at differenttimes--a bi-alphabetism if you wish--of which bilingualism is the fashionableexample. It deserves reemphasizing that the two alphabets map onto the samephonemic and semantic structure.

At this juncture let us collect the preceding discussions of the phonemicregularity and the bi-alphabetism of Serbo-Croatian in order to highlightseveral important contrasts with English orthography. First, where it can beclaimed that the English orthography more directly represents the morphology,it can be claimed that the Serbo-Croatian orthographies more directly repre­sent the phonology. Common to the views of Chomsky and Venezky, a reader ofEnglish often needs to know more about a word than its surface orthographicstructure in order to pronounce it. One would say of Serbo-Croatian thatknowledge about any word's sur face structure is generally allthat is needed in order to pronounce it. Second, English more thanoccasionally reveals the etymology of word but the radical reworking of theSerbo-Croatian accord to Karad s unction ensured thatthe contemporary would be essentially ahistorical. Third, becauseof the virtually invariant relation between letter and sound there are no truehomophones in Serbo-Croatian. ( tuations such as I sewai t/weight could never arise) We em ze truenature of Serbo-Croatian homophones of a very precisely,letter sequences that are visually quite distinct--for one is composed mainlyof uniquely Cyrillic and the other of un y Roman letters--but which areidentical in pronunciation and

allows truerather thanits surface

to differentdifferently and

be short or and itsSometimes these

More commonly, however, thesentential context Theof ain made

sequence such POTOPill ( Table 1), and

inundation and

It is the case, however, that Serbo-Croatian, like English,homographs. It is for this reason that a reader canalways, pronounce a word correctly on the basis oforthography. Two words may be written the same way, but,assignments of vowel and accent can bemean different In Serbo~Croatian a vowel canaccent can or can not ex tend into thecontrasts are noted d marks.ambiguity must be resolved as inlanguage gives rise additionally tomanifest over the two al Thus acan be read one way in Roman and anothermean two entirely d (

208

There is a further feature of the Serbo-Croatian language on which we nowpass remark by way of concluding our delineation of the language's specialpropertiese It is that inflection is the principal grammatical device in thelanguage in contrast with English, which uses inflection for grammaticalpurposes only sparingly. Thus for nouns, all grammatical cases in Serbo­Croatian are formed by add ing to the root form an inflectional element,namely, a suffix consisting of one syllable of the vowel or vowel-consonanttype e The Serbo-Croatian nouns, pronouns, and adjectives are declined inseven cases of singular and seven cases of plural whereas verbs are conjugatedby person and number in six forms.

ERROR IN BEGINNING READING----

Where other languages with a close match between sound and wri ting havebeen examined, the evidence is that children learned very rapidly to readaloud letter sequences congruent with the orthographic rules of the language(Elkonin, 1963; Venezky, 1973). Nevertheless, it can be noted that indiffer­ent to the script-to-utterance correspondence reading differences emerge early(Gibson & Levin, 1975) and that some children will continue to have problemseven where the spelling of the words on which they are instructed isphonetically regular and maps to sound directly (Savin, 1972). Reading skill,in the long run, appears to be largely indifferent to the language being read(Gray, 1956). A not overly venturesome claim is that different writingsystems induce differences in acquisition of reading and differences in theread ing process without necessaril y affecting the ul timate pro ficienc y ofread ing. The point to be emphasi zed perhaps, is that of Carroll (1972): "Aperfectly regular alphabetic may facilitate word-recognition processesbut its use does not alter the fact that the learning of reading entails theacquisi tion of skills in compo word uni ts from their separate graphiccomponents and practice, amounts of it in recognizing particular wordunits."

Given the orthographic distinction between English and Serbo-Croatian onecan ask: In what ways does the beginning reader in Serbo-Croatian differ fromhis counterpart in English and in what ways are they the same? One can ask,in short, with respect to the acquisition of reading, what changes acrossorthographies and what remains invariant? We are examining this question inrelation to research already conducted and currently underway at the HaskinsLaboratories.

ofrom

its users toBut what is required

Liberman and Shankweilercan into

word in his vocabularythe a1

the read research of the Haskinsis somehow on speech. One

awareness" (Mattingly, 1972).But this non anal

the child cannota letter

of the child( 1979) argue

and hecontains and

A ofLaboratories f group is thatrecent focus has been notionA child try to read words

y, while useful to abenefit from the fact that the al

pronunciation from theto know how the althat the child mustmust know how manytheir order He must know

209

phonemes, not syllables or some other uni t of speech (see a1 so Glei tman &Rozin, 1977; Rozin & Gleitman, 1977).

The difficul ty and significance of phonemic segmentation has been fre­quently noted (e.g., Elkonin, 1973: Gibson & Levin, 1975; Rosner & Simon,1971 ); the inability to analyze syllables into phonemes marks the child whohas failed to learn how to read or , at least, who read s poorl y(I. Y. Liberman, Shankweiler, A. M. Liberman, Fowler, & Fischer, 1977; Savin,1972) •

Exemplary of the difficulty with phonemic segmentation is the pattern oferrors a child makes in reading syllables. For simple Englishconsonant-vowel-consonant structures the error rate on the final consonant islarger than that on the initial consonant while the error rate on the vowel islargest of all (Shankweiler & I. Y. Liberman, 1972). Moreover, the form ofthe vowel and consonant errors differ in nontrivial ways (I. Y. Liberman &Shankweiler, 1979). To what extent, one might ask, are these patternings oferrors orthographically based? Are they indigenous to the writing system ofEngl ish or would they be as likely in the orthographies of Serbo-Croatian?For example, the greater error rate on vowels might be owing to the fact thatin English vowel pronunciation is extremely context conditioned. On the otherhand, it might be owing to the differential status of vowels and consonants inthe perception and production of speech; in which case one might treat thedifferent error rates of vowels and consonants and the direction of thedifference as indexing a universal property of phonographic writing systems.

We have begun an examination of these questions through an experimentthat is closely comparable to one previously conceived and conducted by theHaskins Laboratories group.

The 65 subjects in the experiment all tested within the normal range ofintell igence. They were selected from the first grade population of anelementary school system located in Belgrade. Their ages ranged from 6.5 to7.5 years. They had completed their first semester and had an activeknowledge of the Cyrillic al

We devised two lists of the CVC-type monosyllables written in Cyrillic.One hundred CVCs were words and 100 CVCs were pseudowords. The words werefamil iar to fir st graders. In the word and pseudoword lists the 25 Serbo­Croatian consonant phonemes that can occur in both the ini tial and in thefinal positions of a word twice in position. In the majority ofthe trigrams the medial letter was one of the five Serbo-Croatian vowels (Iii,lel,lal lui) in ' , , , , " COK 'juice,' andBYK 'wol f' In some tr med ial letter was the semi-vowelIr/" In Serbo-Croatian mono of the consonant-semivowelI r I-consonant, as in ' , TPH 'thorn rPE' emblem,' are notAnd finally, it should be noted that of the 100 words, 25 could be reversed toprod uce other word. For the word ' 'if read from toleft reads pOB'slave '

A str of threecenter of a

aced face

210

over

the

the examiner. The subject was asked to read each letter string aloud as itwas presented. Responses were written down by the examiner and were recordedsimul taneously on magnetic A complete list was presented in a singlesession with each child in two separate sessions" If in thefirst session the child read the word list, then in the second session he readthe pseudoword list and vice versa The order of presentation was balancedacross children.

The responses to the stimul i revealed several types of errors: 1)substitution, 2) addition, 3) omission, and 4) reversal of sequence when aletter string or a part of it was read from left to right. Single letterorientation errors did not occur because the Cyrillic uppercase letters didnot provide opportunity for rev letter orientation.

The analysis of errors showed that sequence reversals accounted for onlya small proportion of the total of misread letters, al though the lists wereconstructed to provide ample opportunity for the complete reversal of se­quences. (As noted, 25% of the words were "reversible"; and 13% of thepseudowords were words if read from right to left, for example, the pseudowordHHC would become CHH i son' ) "

The complete sequence reversals are distinguished from the partial andthe total reversal scores for words and pseudowords are given in Table 2.Proportions of opportunity for error (in percentages) are presented withinparentheses.. We note that sequence reversals were rare.

Single letter omission errors were also quite rare Their distributionon initial and final consonants and on the medial vowel/semivowel is presentedin Table 3 Omissions of the final consonant in words seem to be morefrequent than in s but the e proportions of opportunityare too small to allow any reliable cone I on distribution.

Additional errors were distributed a nonrandom manner (see Table 4).Additions of a in front of the final consonant (Fe 1) were morefrequent than after the final consonant ( ), other types of additions beingrelatively infrequent.

pseudowords of the consonant-semivowel Irl -consonant type,phoneme in front of the final consonant were relatively

For ex ,the word rPB was often misread- as I grabl ,In four words (rPB f BPX f TPr f TPH) there were 45in four s (BPC, 1 KPTI TIPK) there

Viewed in terms of'!fJOrds the

to 1 Thi is a notablerendition of the letter

ial semivowel and the final

In words andadditions of athe most freI grub/, I

vowel47

itiesto 1

wereoamountsresult"

pernant

in thefinal conso­

Raw error

211

Table 2

Sequence reversals

Complete Partialsequence sequencereversal reversal Total

Words 17% 6 23( 1• 1) (0.0%)

Pseudowords 21 13 34(2.5%) (0.0%)

Table 3

Omission errors

Initial Medial Finalconsonant vowel consonant Total

Words 4 11 16(0.2%)

Pseudowords 4 3 3 10

Table 4

Additions of a single phoneme

Before final After finalInitial Medial consonant consonant

consonant vowel FC1 FC2 Total

Words 6 10 12 80

Pse udo word s 9 52

212

Table 5

Single phoneme substitution errors

Initial Medial Finalconsonant vowel consonant Total

Words 172 93 264 529(2 6%) ( 1 @ 4%) (4 &1%)

Pseudowords 213 113 368 693(3.3%) ( 1 0 7%) (50 7%)

scores and the respective percentages (within parentheses) indicate that finalconsonant (FC) errors exceed initial consonant (IC) errors. A Wilcoxonsigned-rank test on proportions of correct responses revealed that thisdifference was significant (T52=252, ~<Oo001), a result that agrees with thefindings for beginning readers of English. The occurrence of phoneme substi­tutions on medial vowel segments was, however, less frequent than on initial(T53=273, ~<0&001) or final (=202, .001) consonant segments & Serbo-Croatian differs from English: cause more difficulty for beginningreaders than vowels& In an attempt to understand this finding one is remindedthat the vowel set in Serbo-Croatian only five vowels and that theSerbo-Croatian vowels are neatly distinctive in the F1-F2 planeo On thecontrary, within some groups of Serbo-Croatian consonants the distinctive-ness is poor. For ex , within group of four affricates ItSI, ItSj/,IdJ/, /d3j/ the phoneme boundaries are extremely ile. Moreover, in someregions of YugoslaVia the native population aces the voiced affricates/tf/ and Id3/ by their respective voiceless mates ItSj/ and /d)j/o

In our opinion the result of s experiment indicates that the substitu-tion errors (both the in consonant and final consonant) were phoneticallybiased. By far the more errors were the substitutions within thegroup of the All of opportunity forsubstitution in corresponding figuresin the 0

is theTable 5) were more

that even at an earlyis sensi tiveto lex icalical ( ) and

1 Forster & Chambers,

213

DECISION AND PHONOLOGICAL ANALYSIS----It is commonplace to underscore the fact that English spelling is a less

than per feet transcription of the phonology 0 Nevertheless, Engl ish is analphabet in spite of its apparent phonological capriciousness--for eachspelled English word provides strong hints as to its pronunciation 0 Somestudents of reading (eegs, Smith, 1971), however, have felt that the hints areso obscure, the relation between and so opaque, that thefl uent read ing of Engl ish by-passes what must be the complex and ard uousprocess of converting the letter patterns into their related phonologicalformso The idea that the fluent read of English may proceed withoutreference to the phonology is buttressed by the claim that the Englishspelling often preserves morphological relatedness, that is, similar meaning( Chomsky, 1970) 0 Given this claim it is a simple step to supposing that thefluent reading of English proceeds as one might suppose that the fluentreading of logographic wri ting proceeds, that is, wi thout a phonologicalintermediary between the printed word and its meaning (eog., Goodman, 1973).

But forceful arguments can be made and have been made by Rozin andGleitman (1977) to counter these denials of a phonologic strategyo Indeed, asRozin and Gleitman (1977) take pains to point out, the observations question­ing a phonological med iary cut two ways and when looked at carefully addstrength to, rather than weaken, the notion of phonological involvement in thereading of Englishe

It is evident from what has been said about Serbo-Croatian writing, thatnei ther of the two foregoing arguments against a phonological encod ing is

y compelling from the e of that orthographyo Indeed, if anopaque relation between script and and a preserved transcription ofthe morphology are advanced as reasons phonological involvement in thereading of English, then a between script and phonologyand an optimal transcr should be received as reasonsfor phonological involvement Serbo-Croatian.

latenciess

to initiatenot word)

that

At all events, this of the contribution of phonologicalencoding to reading is expression in various laboratorytasks An extremely popular is that of lexical decision, a task in whichthe subject must decide as y as possible whether a visuallyletter string is a worde A finding often as evidence forcal involvement in ish lexical items is that rejectionfor nonhomophonic s are shorter than for homophonic(Rubenstein, Lewis, & Rubenstein, 1971) That is, it takesresponse ( say, indicate "non (it isto a word than todoes not sound any Davelaar, (Jonasson,1977) . While, in the

y mediated access to Meveldt, & Ruddy, 1974), other ex

anal s atto

All things considered, however, the emerging orthodoxy appears to be thatthere is both a phonologically mediated route to the lexicon and a moredirect, nonphonological route with the two modes of access relatively indepen-dent and possibly in As Gleitman and Rozin (1977) expressit, reading probably s at a number of grains of linguistic analysissimul taneousl y 5

the claim of involvement in lexicalthat suggests that in lex ical decision on

s the ical representation cannot be by-ical inter of a letter string is

Additionally, evidence is presented to show aical ev al uation and the lex ical ev al uation of

ficance to the construction of a theory of word

We wish todecision 0 Evidence isSerbo-Croatian letter

and theobligatory and automatic.compl between theletter strings that is ofrecognition 0

Given the nature of and the relation between the two Serbo-Croatianalphabets it is possible to create a variety of types of letter strings.Thus, a letter string composed of un y Roman letters or of uniquelyCyrillic letters (in Figure 1) would receive single phonological interpreta­tion and could be either a word or not a word In contrast, a letter stringcomposed of the common and us letters (see Figure 1) would receive twodistinct phonological inter and could be either a word or not aword; more precisely, it could be a word in one alphabet and a pseudoword inthe other or it could two different words, one in one alphabet andone in the other

In a series of three experiments (Lukatela, Sav ,Gligorijev ,Ognjeno­v & Turvey, 1978) bi-al ects were invited--by experimentaldesign and instruction--to relate to letter (block capitals) in theRoman alphabet mode. None 0 the letter s seen by a subject werecomprised of uni and relativ of the letter

s were composed of common s that is to say, couldeven be read as illic The usion on which all three experimentsconv was that lex ical decision to a letter was slower when thatstring could be en two ical readings (that is could be read ineither the assigned Roman al mode or the nonassigned Cyrillic alphabetmode) but if and if letter was a word in at least one of thealphabetse-- s that could be read in both al were rejected noslower than s constructed from the set of letters unique to theRoman alphabet

215

Table 6. Types of letter strings in the Roman and Cyrillic alphabet.

216

Type ofletter Lexical entry (L) Phonological Symbolic Is it

string representation (P) representation a word?(LS) (in Roman

In In In In or inRoman Cyrillic Roman Cyrillic Cyrillic)(L

R)? (L

C) ? (P

R) ? (PC)?

,<alALSI Yes No Yes No LS1 Yes

o PR

LCLSla No Yes No Yes Yes

Pc

<lA'lCL83 Yes Yes Yes Yes LS3 Yes

PR,PC

LRLS4 Yes No Yes Yes LS4 Yes

'-f) PR, Pc

LR ::: 'CLS5 Yes Yes Yes Yes LS5 Yes

Pc

LS6 No Yes Yes Yes LS6 Yes

"f) 'R, 'c

LS7 No No Yes Yes lS7 No

'f; PR, ·c

LS8 No No Yes No lS8 No

PR-

LS8a No No No Yes lS8a l No

LS9 No No Yes Yes No

'R Pc

and LS9 are composed solely from the common letters (see Figure 1) and aretherefore read the same way and mean the same thing (in the case of LS5) inRoman and Cyrillic. The resul ts of the experiment are shown in Figure 2. Itis apparent from inspection of Figure 2 that lexical decision was impaired forthose letter strings that could be given both Cyrill ic and Roman interpreta­tions but only if the letter string was a word. To give two of the relevantcomparisons, decision times to LS4 were significantl y slower than decisiontimes to LS 1 (F=11. 72; df= 1 ,26; p<O. 01 ); decision times to LS3 were signifi­cantl y slower than deci sion times to LS 1a (F=33. 4; df= 1 ,27; p<O. 001 ) . Thelatter contrast is especially interesting since letter strings of type LS3 arewords in both alphabets and since a general observation in the literature onEnglish words is that letter strings with mul tiple meanings are accepted aswords faster than letter strings with a single meaning (e.g., Jastrzembski &Stanners, 1975). Clearly the present observation is counter to this generalfinding. It should also be noted that the slower decision time to LS3 waswitnessed in our previous research (Lukatela, Savic, Gligorijevic, Ognjenovic& Turvey, 1978). Returning to the data represented by Figure 2, where theletter string was not a word, the lex ical decision was not retarded byphonological bivalence: decision times to LS7 did not differ, for ex ample,from those to LS8(F=2 .. 44, df=1 ,50).

As anticipated, these data on bi-alphabetic lexical decision permit twoconclusions of some significance to an understanding of the reading of Serbo­Croatian .. (We are assuming like others--for example, Coltheart et al .. , 1977-­that lexical decision is a laboratory task well suited to investigating thenature of the information ex tracted from a printed word for use of lex icalaccess.) First, the data strongly that phonological encoding ofSerbo-Croatian words is an automatic and extremely rapid process; as we haveseen, phonological bivalence interferes with lexical decision. Second, thedata suggest that it is not phonological bivalence per se that retards lexicaldecision, rather the necessary contingency is that the phonologically bivalentletter string being evaluated must be a word in the Serbo-Croatian .1

There are a number of theories that could be pursued by way ofthis curious result of bi-alphabetic lexical decision. They are not pursuedhere for there is little to be at this by adjusting the detailsof this or that account of lex ical decision (e.g Of Col theart et al", 1977;Meyer & Rudd y, Note 1) so as to force a fi t wi th the present data" Itsuffices, perhaps, to note the Coltheart et al .. (1977) concluding lament thatfor English there is no compelling evidence for the view that the mapping fromprinted word to lexical entry references the phonology. They propose that:

this view would be obtained demonstrat-code for a word is sometimes used in

to that word in a lex ical decision ora demonstration remains to be achieved

1 ) •

Unequivocal evidence foring that themaking the tv

zation task;( Col theart et al. 9 1

Do the data constitute such a demonstration for Serbo-Croatian?

217

NI--'00

WORDSit ive response)

PSEUDO -WORDS(negative response)

--DOUBLEPHONOLOGY

SINGLEPHONOLOGY

DOUBLEPHONOLOGY

SINGLEPHONOLOGY

2. Lexical decision latencies and errors for Serbo-Croatian letterstrings that are readable in only one alphabet or readable in bothalphabets.

THE PROCESSING RELATION BETWEEN TWO -------A question that has been pursued at some length is how the Roman and

Cyrillic alphabets relate psychologicallYe For the reader of Serbo-Croatianthe alphabets must be kept distinct at some level (or in some manner) ofprocessing in order to circumvent the ambiguous characters asource of confusion. Might we therefore of an alimpl ying that the reader can be in one mode or the other butboth concurrently? The experiments just described bear on this

And ho\'J are the two alphabets memorially If there are twoalphabet spaces are all the letters of the Roman alphabet stored in one spaceand all the letters of the Cyrillic al stored in the other? Or is therea region of overlap, say, the of the common letters? Giventhat the meaning of one alphabet the other, how is inlearning manifest in either the processing or the representation of the twoalphabets? These questions and others guided our attempts to understand thepsychological fi t between the two Serbo-Croatian writing systems (Lukatela,Savio, Ognjenovic, & Turvey, 1978); a part of that research is here.

A very experiment proved exceptionally instructiveo Nativ East-ern Yugoslavians (those who learn Cyrillic first) were presented individualRoman and Cyrillic letters in random order and pressed a key as quickly aspossible in answer to the question "Is this letter illic?" or to thequestion Ii Is this letter Roman?" The resul ts are given in 3" It tookconsiderably longer to verify the common letters (see Figure 1) were Roman inthe II Is this letter Roman?n condition than to ver that the common letterswere Cyrillic in the nIs this letter illic?1i condition. The isthat the ects of the experiment viewed the common letters as essentiallymembers of the ic al and only indirectly as members of the Romanalphabet 0 Arguing in like , the characters would appear toinhabit both al spaces. The most tell observation however was this

lic letters in the Roman al mode took yRoman letter in the mode

relation and( thecharacter

(c,r»learnedcharacterRoman charactersthe av similar

We have come to look at these data in the fol way. We reasonedthat the average latency for rejecting a Cyrillic character as Roman is anindex of the degree to which a description of a Cyrillic character is, on theaverage, similar to a of a Roman character. In the notation ofTver (1977) this similari ty may be written as s( c ,r) where the

of the ic letter (c) is the of thewhere the memorial of an indiv idual Roman letter

a Romantherefore,

\'Jho

219

NNo

750

700

U 650w(/')

~w~

;:600zo.....w«~ 55

500

l----l "IS nus LIHTEIII CV'UU.IC i' 'l'IES

1----4 "1$ THIS LEHEIIIIOUANi' IilO

750

I­

W«..... 550ex

500

----.. ",S THIS UHEIIIIOtIIAM? VEl

I---~ ",S TH'S UHER CYR'LlIC'f' MO

I.."" I...".TII

blmtll2III

.L.

<150

f ! ! • !

AMBIGUOUS COMMON UNIQUElY UNIQUELYLETHRS LETTERS CYRILLIC CYRilLIC

LETTERS LETTERS

45

~ ,AMBIGUOUS COMMON

LETTERS LETTERSUNIOUELYRO"'AN

LETTERS

UNIQUElYROMAN

LETTERS

ure 3 Mean latencies and their range of variation for the alphabetdecision task performed by subjects who learned the Cyrillic alpha­bet first.

from X, the subject term The putative salience of (processing) the Romanalphabet may arise because the dimensions of description of the Roman alphabetinclude those of the Cyrillic; or that the descriptors of the Roman alphabetdistinguish the Roman characters more efficiently than the descriptors of theCyrillic alphabet distinguish Cyrillic characters 0 In short, the basis forthe may lie in some absolute property distinguishing the structureof the two al If true, the direction of the asymmetry should beindifferent to the order in which the alphabets are acquired. On the otherhand, the basi s for the asymmetr y may just be the order 0 f ac qui si tion . Tothis purpose, the alphabet-decision task described above was replicated withsubjects who had acquired the Roman alphabet first and the Cyrillic alphabetsecond" The resul ts are shown in Figure 4. They reveal that under the twoquestion (WI Is this letter Roman?"; "Is this letter Cyrillic?") thesesubjects behaved differently, as did the subjects in the first experimentBut most importantly behavior of the subjects indigenous to WesternYugoslavia was diametrically opposite to that of the subjects indigenous ofEastern Yugoslavia (compare Figure 4 with Figure 3). By the same reasoning asoutlined above we conclude, for subjects who learned the Roman alphabet first,that s(r,c»s(c,r). That is, for Roman-first subjects, processing Romanletters is more similar to Cyrillic letters than vice versa. Moregenerally we conclude that the alphabet-processing asymmetry is owing not to afixed structural property of the alphabets but to their order of acquisition.One tentative conclusion to be drawn is that the procedure developed by thechild to decode the letters of the first acquired alphabet is modified for thesecond acqUired al so that decoding the second acquired alphabetnecessarily entails the procedure for decoding the first acquired alphabet butnot vice versa

But the more outstand ,although equally tentative, conclusionto be drawn is that the order in which the al are , and theconcomitant early bias in read toward one of the al leaves aprofound on the letter decoding processes of adul t readers ofSerbo-Croatian. This conclusion is not unrelated to some resul ts recentlypublished Jackson and McClelland (1979). In the view of some students ofread (g. Kolers, 1969 9 Smith, 1971) individual differences in the readability of experienced readers are solely differences in comprehension abili­ty" The research of Jackson and McClelland brings this view into question byshowing individual differences in the ability of American studentreaders to access letter codes, an ability that accounts for a significant

ion of the variance in effective read What has been noted withmature Serbo-Croatian readers is that in the al decision task there is

interaction between the al first learned and the ald ided upon decision times for Roman-the mirror 0 thefirst surpr about thi is

ects hav read in the two al between 1 and 16 yearon a sian task the al learned st makes its mark"

The point on our data and se of Jackson and McClelland would appearto converge encod processes which letter the

not matureo

221

NNN

~ ··111 TMllI! l.lIliE!! C'III'Hl.liC ..·· '1'£$

1---1 ··11. THill! UUUflilOMAilI?" 160

"Il'~ THUll u,nlUI "OMit"''''" 'I'U

nUl I.lniUI (;'1'11191..1.1<:"'" il&O

iU.1l.aJ

23/J)

!l.aJ

:e

;::

I-

Z

Z

0

0

;::

;::

!2S

u39

u

c:

c(

W

l.aJ

Ill:::

Cl::!13

Mean latencies anddecision performedfirst

range of variation forects who learned the

alRoman al

REFERENCE NOTE

1e Meyer, De E., & Ruddy, M. G. Lexical-memory retrieval based on graphemicand phonemic representations of printed words. Paper presented at themeetings of the Psychonomic Society, St. Louis, 1913.

REFERENCES

ournal ofM. Lexical access and naming time.Behavior, 1913, 621-635.

s In H. Levin & J. P. Williams (Eds.) ,reading. New York: Basic Books, 1910., H. The of Cambridge: The MIT

Baron, Je, & Strawson, C. Use of orthographic and word-specific knowledge inreading words aloud. Journal of Experimental Psychology: HumanPerception and Performance, 1916, --386-393e

Carroll, Je Bo Defining language comprehension: Some speculations. InJ. B. Carroll & Ro Oe Freedle (Eds o), Language comprehension and theacquisition of knowledge. Washington, D.C.: V. H. Winston, 1912.

Chomsky, N. Phonology and reading e In H. Levin & J. Pe Williams (Eds.),Basic studies on New York: Basic Books, 1910.

Coltheart, Jonasson, J., & Besner, D. Access to theinternal lexicon In Se Dornic (Ed e), and performance VIeLondon: Academic Press, 1911.

Elkonin, D. Be The psychology of mastering the elements of read ing. InB. Simon & J Simon (Eds.), Educational psychology in the U.S.S.R.London: Routledge & Kegan Paul, 1963.

Elkonin, D. Be U.SeS.R. In Jo Downing (Ed.), Comparative reading. New York:Macmillan, 1913.

Forster K. K., & Chambers, S.Verbal ----=

Francis, W. N.studies

Gibson,Press, 1975

Gibson, E. J 0' Pick, A. , Osser, H., & Hammond, Me The role of grapheme­phoneme correspondence in the perception of words. American ournal ofPsychology, 1962, 15, 554-510.

Gleitman, L. R. , & Rozin, Po The structure and acquisition of reading I:Relations between orthographies and the structure of language. InA S Reber & D. L. Scarborough (Ed s.) , Toward a psychology of reading:The of the CUNY Hill sdale , N. J 0=-- LawrenceErlbaum , 19110 --

Goodman, Ko So The 13th easy way to make learning to read difficul t: Areaction to Gleitman and Rozin Research 1913, li,

an

223

Sutherland (Ed ) ,N J.: Lavlrence Erlbaum Associ-

visual access in read

of, W. S.Pari

Green , D. W. & Shall ice, T.&_~~__ t 1

&

The

and

Wien:

Ko S. Goodmanof

On the relationA preliminary

~__ , 1978,

An initial

In

P. , & Turvey, M. T.~__ , 1978, ~, 142-

ournal of

M G. Loci of contextual effectsDornic & P. M. A. Rabbitt (Eds.),Academic Press, 1974.,d ia and the phonological

ournal of

Bo, Ognj enov ic ,

jezika po govoru prostoga naroda.

RuddyIn S.

London:. J.

D., Liberman, A. M. Fowler, C., & Fischer,and recod ing in the beg reader 0 In

(Eds.) a of The..:::.--"'----="- - --

CUNY conference 0 Hillsdale, No J. : Lawrence Erlbaum

M., GI j evlex ical decision.

, Mo, Ognjenov , Po, & Turvey, M. T.the Roman and Cyrillic alphabets:

readers and

Say

wi th bi-al

DelY., & Shankweiler,

In L. BResnick & P. A ~leaver

) Hillsdale, N. J

Matt

Lukatela, G ,Bi-al165

Lukatela, Gobetweenanal11

1Liberman, I. Y., Shankweiler,

F 0 W0 PhoneticAo Sa Reber & D.

of the, 1

SaY

KolersJ.Newark,

Liberman, Iread

Jastrzembski, J. Eo, & Stanners, R. Multiple-word meanings and lexical searchspeed. of and 1975, 534-537.

Karad V. S. Pismenica serbJ. Sc hn irer Pr e s s, 181 4 .

Kleiman, G. M recoding in reading.4only incidentally visual.

and the

224

to

and

in word

and third1973, 64,

1977, .§!±.,Hague: Mouton,

words: An1-150

A psycholinguisticHol t t 1971 Q

P

letter-sound izationsldren. ournal of

Massaro, D. W The role of orthographicIn Lo B. Resnic & P. A. Weaver (Eds ),

(Vol 1). Hillsd t N J. Lawrence

Vene

Vene

Press, 1972, 293-317 $

Smith, F. Understanding ----=to read. New Yor .

-::---~-='--- 0 f similarLe The structure of

TverVenezky, R@

1970 Q

R. L Thee Finnish

analthis vol

bivalent letter

these data (see Lukatela, Po padthe detriment to performance incurred

s occ urred both for word sand

225