Sunteți pe pagina 1din 28

This article was downloaded by: [Universidad Autonoma del Estado de Mexico]

On: 8 October 2010


Access details: Access Details: [subscription number 906963981]
Publisher Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

The Quarterly Journal of Experimental Psychology Section A


Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713683590

Bilingual lexical representation: The status of spanish-english cognates


Paula Cristoffaninia; Kim Kirsnera; Dan Milecha
a
University of Western Australia, Perth, Australia

To cite this Article Cristoffanini, Paula , Kirsner, Kim and Milech, Dan(1986) 'Bilingual lexical representation: The status
of spanish-english cognates', The Quarterly Journal of Experimental Psychology Section A, 38: 3, 367 — 393
To link to this Article: DOI: 10.1080/14640748608401604
URL: http://dx.doi.org/10.1080/14640748608401604

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
The Quarterly Journal of Experimental Psychology (1986) 38A,367-393

Bilingual Lexical Representation:


The Status of Spanish-English Cognates
Paula Cristoffanini,
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

Kim Kirsner and Dan Milech


University of Western Australia, Perth, Australia

Two experiments were conducted to determine the functional status of


cognates. Two hypotheses were considered. According to the first hypo-
thesis, language is a critical feature governing lexical organization, and
cognates may therefore be equated with morphologically unrelated trans-
lations. According to a second hypothesis, however, language is not a
critical feature governing lexical organization. Instead, the boundaries
between perceptual categories are determined by morphological consider-
ations, and cognates may therefore be equated with intra-lingual varia-
tions such as inflections and derivations. If the first hypothesis is correct,
cognate performance should follow that observed for translations, but if
the second hypothesis is correct, cognate performance should follow that
observed for inflections and derivations.
The experiments used different procedures in order to discount task-
specific explanations. The first experiment involved repetition priming in
a lexical decision task, and emphasis was placed on relative priming; that
is, on the amount of facilitation which occurs when, for example,
OBEDIENCIA primes OBEDIENCE, expressed as a fraction of the
amount of facilitation that occurs when the same word is presented on
each occasion (i.e., when OBEDIENCE is used to prime OBEDIENCE).
The second experiment tested memory for language. Four types of
cognates were tested. These were: orthographically identical cognates,
regular cognates with cionltion substitution, regular cognates with dad/ty
substitution, and irregularly derived cognates.
The results were unequivocal. The priming values observed previously
for cognates were qualitatively and quantitatively similar to those observed
for inflections and derivations, and this classification was confirmed in the
second experiment, involving memory for language. The results are
consistent with the general proposition that morphology rather than

Requests for reprints should be sent to Kim Kirsner, Department of Psychology,


University of Western Australia, Nedlands, Perth 6009, Australia.
The authors would like to thank Stephen Monsell for his helpful comments on an
eatlier draft of the manuscript.

0 1986 The Experimental Psychology Society


EP c) 38/3-B
368 P. Cristoffanini et al.

language governs the boundaries between perceptual categories, and a


number of specific explanations are reviewed.

GENERAL INTRODUCTION
This paper is concerned with the psychological concomitants of
bilingualism. Of particular interest is the extent to which the processes
and structures that underlie language comprehension and memory are
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

specific to the individual languages of a bilingual.


The general issue of specificity has been approached in a variety of
ways, depending on the type of process or representation under
investigation. McCormack (1976, 1977), for example, approached the
problem from the point of view of a memory theorist, with reference to
memory for the occurrence of particular words. McCormack concluded
that language-in the bilingual sense of the term-may be compared
with other “attributes of memory” (Underwood, 1969) such as case,
speaker’s voice, spatial position and orthography, and he gave it no
particular prominence. Kolers (1978, 1979), on the other hand, gives a
language a crucial or defining role in the “representation of experience”.
According to Kolers, knowledge is generally organized in terms of the
means or procedures by which it is acquired, and, as language in
particular is classified as a “means” or “procedure”, it follows that the
organization of knowledge is language-dependent.
Albert and Obler (1978) and Perecman (1984) have developed a more
flexible position in which they argue that the degree and the form of
language specificity depends on the type of processing concerned, and
they describe a provisional frame of reference involving a series of
cognitive levels. In keeping with the view that the form of the
relationship between the representations and processes that underlie
bilingual performance can be specified only on a function-by-function
basis, the experiments reported here selectively address lexical function,
and inference is restricted to that domain.
Further distinctions may be required, however, even within the
lexical domain. One such distinction involves the contrast between
processes that are engaged by individual words and those, presumably
additional, processes that are engaged when pairs or sets of words are
presented. Where pairs of words are concerned, it now seems likely that
some of the relevant processes are language-independent. For example,
the structures derived from multi-dimensional scaling of word ratings
appear to be language-independent (Barnett, 1977). Similarly, although
it is not clear whether semantic priming reflects access to intensional or
extensional processes, lexical decision tasks involving pairs of words
from different languages yield semantic priming effects that are difficult
Bilingual Lexical Representation 369

to distinguish from those obtained under monolingual conditions


(Kirsner, Smith, Lockhart, King, and Jain, 1984, Experiment 4; Meyer
and Ruddy, 1974). Furthermore, the decay characteristics for semantic
priming are at least qualitatively similar under inter- and intra-lingual
conditions (Kirsner et al., 1984, Experiment 5).
Where individual word processing is concerned, however, the effect
of language is quite different. Classification of a word in one language is
not facilitated by prior exposure to its translation (Kirsner, Brown,
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

Abrol, Chaddha and Sharma, 1980; Kirsner et al., 1984; Scarborough,


Gerard and Cortese, 1984). That such words are treated as if they are
new is consistent with the hypothesis that the perceptual categories (and,
by implication, the lexical units) underlying verbal function are
language-specific. However, given the types of translations that have so
far been studied, the absence of inter-lingual transfer in repetition
priming is open to two interpretations. The first possibility is that it is
the language change per se that is critical. Provided that language is
unambiguously indicated by either contextual or orthographic cues,
translations will be treated as if they belong to different perceptual
categories, and repetition priming will not, therefore, be observed.
According to the alternative interpretation, however, the language
change is irrelevant. The critical feature is morphology, or word
structure. Where translations share some morphological elements,
inter-lingual transfer will be observed, but where they do not,
inter-lingual transfer will be absent.
Because it is possible to vary their morphological and lingual
properties independently, cognates provide a crucial test of the above
hypotheses. If language is an over-riding feature, acting as a filter that
determines the limit of each perceptual category, performance on
OBEDIENCE will not benefit from prior exposure to OBEDIENCIA,
despite their morphological similarity. However, if lexical organization
is governed by morphology rather than language, performance on
OBEDIENCE may benefit from prior exposure to OBEDIENCIA, and
the magnitude of this benefit should be comparable to that observed for
intra-lingual word pairs that share structural properties, such as
inflections and derivations.
Evidence from two areas of research suggests that morphological
principles transcend language barriers under certain circumstances. The
first of these is polyglot aphasia, where it is evident that symptoms such
as morpheme blending and morpheme shifting can occur across
languages just as they do within languages in monolingual aphasia and
slips of the tongue (Perecman, 1984).
The second body of evidence is closer to the methodology of the
present paper. Caramazza and Brones (1979) found that lexical decision
370 P.Cristoffanini et al.

times for cognates were determined by their status in the subject’s


preferred language, Spanish. When the words (e.g., GENERAL-
GENERAL) were presented in an English context, mean reaction-time
was faster for the cognates than for the control words, but the relevant
means for the cognates and control words were virtually identical when
the same words were presented in a Spanish context. This demonstration
is directly relevant to the present argument. If effective frequency for
each cognatic form is determined by the frequency of occurrence of the
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

preferred language form, it follows that a shared or common


representation is normally accessed when each of the relevant forms is
presented; that is, that morphology rather than language determines
the functional limits of each perceptual category. T h e experiments
reported here were conducted to test the generality of this hypothesis,
using different experimental techniques and a variety of cognatic
forms.

The Perceptual Record, Repetition Priming


and Memory for Stimulus Class
In order to discount task-specific explanations, the experiments
reported in this paper used different paradigms. The selected paradigms
were repetition priming in a lexical decision task and memory for
stimulus class, respectively. The advantage of such a combination is that
if they yield complementary results, the combination provides a measure
of protection against the problem of indeterminacy (Anderson, 1978).
According to Pylyshyn (Note 1, cited by Anderson, 1978),
The appropriatesubject of our analysis of representation should be not the
representation per se but a representational system consisting of the pair
(representation,process).
Our response to the problem of indeterminacy is different from that of
Anderson (1978), however. Whereas he argues that it is impossible to
evaluate a representational claim unless the processes that operate on
that representation are also specified, our approach relies on the use of
different tasks to isolate the representational contribution. There is one
important assumption, however. This assumption is that lexical
decisions and memory decisions about stimulus class involve different
processes or algorithms. If this assumption is valid, and if the
relationship between the treatment effects (i.e., cognate classes) is
task-independent, that pattern may be attributed to the representational
contribution. If the assumption is invalid, however, then the pattern
may stem from either source.
The combination of tasks used in this paper is based on the
assumption that a record of perceptual processing is implicated in each
case (Kirsner and Dunn, 1985). The central element in the argument is
Bilingual Lexical Representation 371

that stimulus identification defines a record consisting of the processes


invoked to solve the problem. This record is then available during
subsequent processing and may contribute directly to performance, as
an activated pathway, or indirectly, as a product of that activation. This
distinction parallels that developed by Tulving (1983). The record may
be preserved in the form of a change in the skill or facility with which the
same or similar problems are solved in future, or in the form of
“declarative” knowledge about the original event.
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

For repetition priming, the extent to which the same processes are
invoked during the study and test phases of the experiment determines
the level of performance. Where the intersection between these sets of
processes is high, substantial facilitation will be observed; but where it is
low, unique processes will be engaged on each occasion, and little or no
facilitation will be observed. In the case of memory for stimulus class it
is assumed that decisions are based on a comparison between the test
stimulus and the available record of the old stimulus, presented during
the study phase. As discrimination is required in this task, increasing the
intersection between the processes invoked on the two occasions will
impair performance.
The proposition that a common principle is involved in repetition
priming and memory for stimulus class is based on meta-analysis of the
published data in each domain. The proportion of transfer observed in
repetition priming and the reliability of memory for stimulus class are
negatively correlated. At the extremes, for example, case and speaker’s
voice yield virtually complete transfer (Scarborough, Cortese and
Scarborough, 1977; Jackson and Morton, 1984) but poor memory for
stimulus class (Kirsner, 1973; Craik and Kirsner, 1974), whereas
language yields no transfer (Kirsner et al., 1980) but reliable memory for
stimulus class (Kintsch, 1970; Kirsner and DUM, 1985). The
limitations of meta-analysis are clear, however, particularly where
correlation is concerned (Brown and Kirsner, 1980), and the present
application permits an analysis of the effect of translation class on this
relationship within the confines of a single study.

EXPERIMENT 1
The first experiment involved an analysis of repetition effects in lexical
decision. In the first or study phase, subjects were presented with
lingually homogeneous lists of Spanish and English words, and asked to
name each word. In the second or test phase of the experiment the
English words were re-presented in English (treatment EE), the Spanish
words were re-presented in English (SE), and a comparable list of
English words were presented for the first time (NE).
Five translation classes were tested. The first class consisted of
372 P.Cristoffanini et al.

orthographically identical cognates such as those used by Caramazza


and Brones (1979). The second and third classes consisted of cognates
with a regular rule which can be used to transform one member of the
pair into the other member. The fourth class consisted of irregular
cognates, where more than one derivational rule is required, and the
fifth class consisted of morphologically unrelated translations. The
classes were:
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

1 . Orthographically Identical Cognates. The English and Spanish


forms of these words were graphemically identical (e.g., REUNION-
REUNION, FESTIVAL-FESTIVAL).

2. Common stem cognates with regular suffix cion-tion. The English and
Spanish forms of these words were orthographically identical except for
their endings. The Spanish forms ended in cion and the English
forms ended in tion (e.g., OBSERVACION-OBSERVATION,
SUCCION-SUCTION).
3. Common stem cognates with regular sufix dad-ty. The English
and Spanish forms of these words were also identical except for
their endings. The Spanish forms ended in dad and the English
forms ended in ty (e.g., CRUELDAD-CRUELTY, HOSTILIDAD-
HOSTILITY).

4 . Common stem cognates with irregular suffixes. The English and


Spanish forms of these words were identical except that (a) the last one,
two, or three letters were different, and (b) the mapping between the
English and Spanish forms was not consistent across all items with a
particular ending (e.g. CALUMNIA-CALAMITY, ITINERARIO-
ITINERARY).
5 . Morphologically unrelated translations. The English and Spanish
forms of these words bore neither orthographic nor phonological
resemblance to one another. These words acted as control words (e.g.
PANADERIA-BAKERY, TRISTEZA-SADNESS).

The argument developed in the introduction may now be used to set


out three models and their associated predictions. According to the first
account, language is an overriding feature. When there is a change in
this feature, a different lexical representation is necessarily accessed. If
this account is correct, repetition priming will be restricted to the
intra-lingual conditions. Although some caution may be required where
orthographically identical cognates are concerned, because their lingual
status depends entirely on context, there should be no facilitation in the
other cognate repetition conditions.
Bilingual Lexical Representation 373

According to the second account, morphology rather than language


governs lexical representation, and reference may therefore be made to
general theories of word recognition. The decomposition model
developed by Taft and Forster (1975) provides a limiting case.
According to their theory, word recognition involves two stages. The
first is a pre-lexical stage, in which affixes are identified and removed,
presumably on the basis of orthography. The second is a search stage, in
which a representation corresponding to the stem is matched with
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

similarly circumscribed entries in the lexicon. The implication for


repetition priming is clear. As the critical lexical entry is accessed by
derivations and by the base word, derivation-to-base priming should be
equal to base-to-base priming. The implication for cognates is also clear.
Provided that phonological differences between the two forms are not
critical, and there is some evidence to suggest that this is the case where
derivations are concerned (Bradley, 1979; Downie, Milech and Kirsner,
1985), cognate-to-cognate priming should be equal to intra-lingual
priming. For example, in the case of PUBLICIDAD and PUBLICITY,
the suffix will be removed during the pre-lexical stage, a common
language-independent representation will be accessed, and repetition
priming should be evident when the same sequence of operations is
applied to the cognate.
The principle of decomposition may be refined and applied in a
number of ways, however, as Taft and Forster (1975) noted. One such
refinement involves the concept of delimitation (Henderson, 1984)
where the basic idea of morphemic representation is preserved, but
access is strictly limited to, for example, regular inflections and
derivations. The intra-lingual evidence on this question is mixed (Hen-
derson, 1985; Kempley and Morton, 1982; Stanners, Neiser, Hernon and
Hall, 1979), but if the hypothesis is valid, and if it extends to cognate as
well as derivational relationships, facilitation should be restricted to
regular derivations; that is, to the first three classes specified above.
The third account dispenses with the notion of lexical entries and unit
delimitation altogether. I n keeping with a line of argument initiated by
Posner (1969) and developed by Kolers and Roediger (1984), we
propose that: (1) stimulus recognition involves a sequence of operations
or processes, (2) a record of the processes invoked to solve each problem
is preserved, in either procedural or declarative form, and ( 3 )
intersection among the processes involved during encoding and testing
determines the extent of relative priming (RP), where this is defined as
follows:
RP = (alternative form - control) / (same form - control)
One implication of the third hypothesis is that if the set of critical
features is large, and the contribution of each one is small, model-fitting
374 P. Cristoffanini et al.

procedures may be more appropriate than inferential statistics. Thus,


priming should be considered in relative rather than in absolute terms.
The repetition priming literature is already replete with Type 2 errors,
and the variability associated with the instruments available is such that
only two or three statistically different RP levels can be identified over
the full range.
As an illustration of the advantages that may emanate from the use of
the RP statistic, and in order to provide a benchmark for cognates,
Stanners and colleagues’ (1979)data was re-analysed in terms of this
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

variable. Table I summarizes the data from nine conditions in their


experiments, identifying the nature of each treatment, and showing the
absolute and relative priming values. Analysis of the RP effects in their
data suggests that one variable-letter differences between the stem of
the inflection or derivation and the base word-accounts for a
substantial proportion of the variance among their treatment means.
Averaged across experiments, the mean RP value for their no change,
small(er) change, and great(er) change treatments are 0.99 (range = 0.94
to 1.09), 0.60 (0.58 to 0.62), and 0.33 (0.27to 0.39), respectively.
T o summarize, two general positions may be identified. According to
the first of these, language is a critical feature governing unit
delimitation, and transfer between the members of cognate pairs should
therefore be absent. The first experiment was designed to test this
proposition using the repetition paradigm. The second position is that
language is not a critical feature, but, depending on assumptions about
the role of morphology in lexical access, a variety of specific predictions
may be made. One possibility involves the principle of decomposition;
the prediction being that magnitude of priming will be the same under
inter-lingual and intra-lingual conditions. A further refinement of this
position involves the concept of regularity, with transfer being restricted
to regular cognates. Another possibility involves the concept of a
perceptual record, from which three deductions may be made. First,
study-test variations in each of a number of stimulus features will
reduce RP. Second, RP will decrease as the proportion of altered
features increases. Third, RP and memory for stimulus class will be
negatively correlated.

Method
Design
The independent variables were repetition status, EE, SE, and NE, and
translation class (as described above), and a repeated measures design was used.
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

Summary of Priming Effectsfrom Four Experiments Involving Morphological Variation between Prime and Test Forms
Repetition Treatment

Experiment and Alternative Same Relative


Type of Variation Example DV New Form Form Priming

Inflections (s) pours-pour PC 95.8 98.8 99.0


RT 932 751 766 1.09
Inflections (ed) burned-burn PC 92.9 98.5 98.3
RT 942 792 782 0.94
Inflections (ing) lifting-lift PC 98.8 100.0 98.8
RT 904 773 764 0.94
Lesser change in stem of hung-hang PC 92.0 99.5 98.0
irregular verbs (1 letter) RT 783 734 699 0.58
Greater change in stem of shook-shake PC 98.0 99.0 99.5
irregular verbs (1 letters)
+ RT 803 764 704 0.39
Lesser change in adjectival selective-select PC 97.0 98.5 97.5
stems RT 899 827 779 0.60
Greater change in adjectival descriptive-describe PC 95.5 95.0 99.0
stems RT 924 892 806 0.27
Lesser change in verb stem appearance-appear PC 89.0 96.0 98.0
RT 926 843 792 0.62
Greater change in verb stem destruction-destroy PC 95.0 96.0 98.0
RT 939 897 815 0.34
(Stanners, Neiser, Hernon and Hall, 1979). DV = Dependent Variable. PC = percent correct. R T = reaction time (msec).
376 I?. Cristoffanini et al.

Subjects
Eighteen Spanish-English bilinguals participated in the experiment. Their
mean age was 32.6 years (standard deviation= 9.3), and there were ten males and
eight females. Eleven of the subjects-five males and six females-were native
Latin Americans, the other subjects-five males and two females-were native
Spaniards. All subjects claimed fluency in both Spanish and English, and their
use of the languages extended to at least five years.

Stimuli
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

The Collins Spanish-English, English-Spanish dictionary (Smith, Bermejo and


Chang-Rodriguez, 1971) was used for stimulus selection, and the items were
chosen so as to fulfil the following requirements: (1) the words had to belong to
one of the five translation classes defied above, (2)all words had to be singular
nouns, (3) compound words were excluded, (4)the words did not constitute
homographs in either language, and ( 5 ) all of the words had a frequency of usage
in English of less than 30 per million, as assessed by Kucera and Francis (1967).
Within this range, the words of each translation class were chosen so that their
frequencies matched the frequency of the words in every other list. The mean
frequency value of the items in each list was 4.65 per million, with a standard
deviation of 5.49.The number of syllables per word varied from two to six.
Seventy-five English-Spanish word pairs that met the above criteria in each
of the translation classes were chosen for testing. Thus, with five translation
classes, there were a total of 375 word-pairs. Three word arrangements were
prepared so as to permit all of the words to occur in each of the priming
conditions, and subjects were randomly allocated to each of these conditions. In
addition to the word set, a further 190 non-words with legal English spelling
were constructed. These non-words were constructed so that 38 examples
resembled the words in each of the five translation classes with regard to their
endings. The non-words were also constructed to resemble the “real” words
with respect to length.

Procedure
The experiment was conducted in two phases in a single session; the first and
second phases involving priming and testing, respectively. Subjects were not
informed that the experiment was to be conducted in two phases, or that words
presented in the first phase would be repeated later in the experiment.
The stimuli were displayed on a BWD cathode ray oscilloscope with an
effective screen area of 10.0 cmz and a resolution of 1024 x 1024 points. The
stimuli were presented in capital letters to obviate the need for stress marks in
Spanish. The letters were about 12 mm high. Subjects responded on a multiple
response box with buttons marked YES and NO. Reaction time and accuracy
data were recorded. A PDP 11/10computer was used to randomly select and
display the stimuli and to collect the responses.

Phase 1. Subjects were instructed to name each word as it was presented on


the oscilloscope. They were also informed that their reading responses would be
recorded, that there would be two blocks of words, one per language, and which
order the languages were to be presented in.
Subjects were presented with 250 words in the first phase of the experiment,
Bilingual Lexical Representation 377

125 in English and 125 in Spanish. Within each set of 125 words, 25 words from
each translation class were randomly presented, one at a time.

Phase 2. Subjects were informed that they would be presented with a mixed
sequence of genuine and false English words, and that their task was to classify
each item as “genuine” or “false”. They were instructed to respond to each
word by pressing one of two buttons on a keyboard-a YES button for genuine
words and a NO button for false words. Ten practice trials were given, and
subjects were instructed to respond as quickly as possible.
Subjects were presented with a total of 565 stimuli, 375 genuine English
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

words and 190 non-words. There were 25 genuine words from each of the 15
treatments given by the factorial combination of repetition status (3) and
translation class ( 5 ) .
As in Phase 1, the words appeared on the oscilloscope one at a time, in
random order. Each word was displayed for 3 sec or until the subject responded
(whichever was shorter), when the next trial was initiated. If subjects failed to
respond within 3 sec, an error was recorded, and the sequence continued.
The first and second phases of the experiment lasted approximately 15 and 30
min, respectively, and they were separated by a 10-min interval during which
autobiographical details were recorded.

Results and Discussion


T h e naming responses made during the first phase of the experiment
were mostly correct. Out of a total of 4500 responses, only seven were
pronounced incorrectly, and six of these were to words from the
orthographically identical cognate condition. The seventh error was the
result of misreading.

Repetition Priming
Table I1 summarizes the results of the first experiment. Substantial
repetition priming was observed in all of the intra-lingual conditions,
and, to a lesser extent, in the inter-lingual conditions, and the effects are
manifest in both the reaction time and accuracy data. Broadly, the
results are consistent with the proposition that the perceptual categories
involved in lexical decision are sensitive to morphology but not to
language. As previously reported, priming was absent in the condition
involving morphologically unrelated translations, but substantial or
complete priming was observed in each of the four cognate conditions.
As shown, the RP values ranged from 0.74 to 0.98.
A two-factor analysis of variance performed on the data demonstrated
that the interaction between repetition status and translation class was
significant, F(8, 135) = 2.13, MSE 4289, p < 0.05. Further comparisons,
L=

using Newman-Keuls, demonstrated that: (a) significant facilitation


occurred under EE conditions for all five translation classes, (h)
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

Table I1
Experiment 1 :Lexical Decision. Mean Reaction Time and Accuracy as a Function of Translation Class and Repetition Treatment.

Repetition Treatment

Translation Alternative Same Relative


Class Example DV New Form Form Priming
~

1 Orthographically identical festival-festival PC 97.1 98.0 98.2


cognates RT 799 745 733 0.82
2 Cognates with regular suffix: observacion-observation PC 96.9 98.0 98.0
cionltion RT 921 815 813 0.98
3 Cognates with regular suffur: publicidad+ublicity PC 94.2 97.1 97.8
dadity RT 866 796 783 0.84
4 Cognates with irregular suffixes disciplina-discipline PC 93.1 96.4 98.2
RT 858 792 769 0.74
5 Morphologically unrelated panaderia-bakery PC 92.4 91.3 96.9
translations RT 908 906 81 1 0.02

DV = Dependent Variable. PC = percent correct. R T = reaction time (msec).


Bilingual Lexical Representation 379

significant facilitation occurred under SE conditions for the four cognate


conditions only, and (c) there were no significant differences between the
EE and SE means in the cognate conditions.
The main effect of translation class was also significant, F(4,
68) = 1 1.79, M S E = 9002, p < 0.001. Newman-Keuls tests revealed,
firstly, that the mean reaction time for orthographically identical
cognates was faster than the reaction time recorded in the other
conditions, and, secondly, that mean reaction time for non-cognates was
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

slower than the other conditions-except for the regular suffix: cion-tion
condition. The means for the five conditions were 759 msec (97.8%
correct), 850 (97.6%), 815 (96.4%), 806 (95.9%) and 875 (93.5%) for
orthographically identical cognates, cognates with regular suffix:
cion-tion, cognates with regular suffix: dad-ty, cognates with irregular
suffixes, and morphologically unrelated translations, respectively. The
main effect of repetition status was also significant, F(2, 34) = 54.8
M S E = 3347, p < 0.001.
The accuracy effects were generally consistent with the reaction time
data, and there was no suggestion of a speed-accuracy trade-off across
conditions. As was the case in the reaction time analysis, intra-lingual
facilitation was virtually complete in the case of cognates and eliminated
altogether in the non-cognate condition. An analysis of variance
indicated that the interaction between repetition status and translation
class was significant, F(8, 136)=2.57, M S E = 14, p<0.05. The main
effects of translation class, F(2, 68)=8.36, M S E = 19, p<O.OOl, and
repetition status, F(2, 34) = 9.20, M S E = 23, p < 0.001, were also
significant.

Frequency
Following Caramazza and Brones (1979), the status of language as a
critical feature governing unit delimitation may also be tested by
examining frequency effects. Indeed, if the same mechanism governs
both frequency and repetition effects (Monsell, 1985), the results of
these tests should be complementary. If interlingual transfer occurs
between cognates in the repetition task, the efficiency with which
English words of a given frequency are classified should be sensitive to
the frequency of their Spanish cognates. Although the first experiment
was not specifically designed to test this prediction, a valid test may be
conducted provided that consideration is restricted to (a) appropriately
selected stimuli and (b) data obtained under new test conditions.
The frequency analysis should provide confirmatory evidence that
language is not a critical feature. Assuming that frequency and repetition
effects involve a common mechanism, it follows that the efficiency with
380 P. Cristoffanini et al.

which new English words are classified will be sensitive to the frequency
of occurrence of their Spanish cognates.
In order to test the frequency hypothesis, pairs of English words were
selected from the first four translation classes such that: (a) both words
had the same English word frequency, and (b) the Spanish word
frequency of the cognates differed, one being low and the other high.
Forty-seven pairs were found: the mean English frequency of these
words was 4.2 per million (Kucera and Francis, 1967); and the mean
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

frequencies for the corresponding Spanish cognates were 0 per million


and 14.7per million (Juilland and Chang-Rodriguez, 1964)for the low-
and high-frequency sets, respectively. DECADENCE and DONATION
are an example of such a pair. Each word has an English word frequency
of 2 per million but their Spanish cognates have quite different word
frequencies; 18 per million for DECADENCIA and 0 per million for
DONACION.
The results for the entire set of 47 pairs show the expected frequency
effects. With inclusion restricted to data obtained under new (not
primed) conditions, the mean reaction times were 888 and 784 msec for
English words with low- and high-frequency Spanish cognates,
respectively; and this difference was significant, F(1, 46)= 10.6,
M S E = 24127,p < 0.001.
The implications of the repetition and frequency results are
considered in the General Discussion.
Translation Class
The translation class differences shown in Table 11 may be explained in
several ways. One possibility is that they reflect difference in word
length, where the mean values for the five classes shown in Table I1 were
7.2, 10.8,9.1,8.3,and 7.6 letters, in that order. A second possibility is
that they stem from the peculiar frequency effects to which bilinguals
are subjected. If it is assumed, for example, that our subjects have been
exposed to English and Spanish words equally often, and that the
effective frequency count for each cognate pair is equal to the sum of
their monolingual frequency counts divided by two, the frequency
matching achieved prior to experimentation is distorted to some extent.
In fact, the new values for the five classes shown in Table I1 are 4.5,5.6,
4.5, 6.2,and 2.4per million, respectively. A third possibility is that the
translation class differences reflect an as yet unidentified morphological
or orthographic property.
Regression analyses were conducted to examine the first and second of
the above hypotheses. Three variables-word length, derived word
frequency (as illustrated above), and reaction time-were included, and
each variable was predicted from the other two. The analyses were
Bilingual Lexical Representation 381

performed on the subject values for each treatment class. The


hypothesis that the proportion of variance accounted for was equal to
zero could be rejected at alpha = 0.05 for each variable, but the actual
proportion of variance accounted for was less than 8% in each case.
None of the hypotheses identified above can be discounted.

EXPERIMENT 2
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

The second experiment involved a test of incidental memory for


stimulus class: language, in this case. In the first phase of the experiment
subjects were presented with a mixed list of English and Spanish words,
on cards, and instructed to read through the list at their own pace. In the
second phase, all of these words were re-presented, in either the same
language or in the alternative language, and subjects were instructed to
classify each word as “sameyyor “different” with reference to language
of presentation on the two occasions. As the experimental task is
meaningless where orthographically identical cognates are concerned,
this condition was exempted, but items from the other four translation
classes were included.
The assumption underlying the second experiment is that the limits
of the perceptual categories available to the recognition system may be
inferred from explicit memory tests as well as repetition priming. The
circumstances under which this assumption is valid are limited. If
learning is intentional, or if words are presented in discourse, the range
of processing options open to people increases dramatically (e.g., Fisher
and Cuervo, 1981), and performance will be dominated by a different set
of variables. But the impact of these conceptually driven processes will
be minimized in an incidental learning task involving isolated words,
and the ease with which the original stimulus class can be reported under
these circumstances should therefore reflect the extent to which the
relevant forms involve different representations.
Where cognates are concerned, prediction is straightforward. If
language is a critical feature, guaranteeing distinct representations
where different languages are involved, memory for language will be just
as reliable for cognates as it is for morphologically unrelated
translations. But if language is irrelevant, and morphological
considerations govern lexical representation, memory for stimulus class
will be less reliable for cognates than it is for translations.
A number of experiments involving either translations or derivations
provide an appropriate basis for quantitative prediction. For four
experiments that used isolated translations, incidental learning
conditions, and comparable stimulus numbers and intervals to the
present study, mean accuracy was 0.85 (standard deviation = 0.06); the
382 P.Cristoffanini et al.

experiments being published by: Kirsner and Dunn (1985); Rose and
Carroll (1974); Saegert, Hamayan and Ahmar (1975); and Winograd,
Cohen and Barresi (1976). If language is a critical feature, performance
in the cognate conditions should approach this figure. However, if
lexical representation is governed by morphological considerations
without reference to language, performance should be inferior to this in
the cognate conditions, although the actual level is difficult to predict on
theoretical grounds. One possibility is that cognates will behave like
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

“surface” features such as case and speaker’s voice, where accuracy is


generally about 60% correct under incidental learning conditions
(Kirsner and Dunn, 1985). Another possibility involves consideration of
memory for derivational status. In a recent study, Downie et al. (1985)
used a mixed list of base words and derivations and found that memory
for derivational status ranged from 56.0 to 64.0% correct for four
conditions that involved systematic variation in pronunciation
differences between base words and the stems of their derivations.
Specific qualitative predictions may also be made about the effects of
decomposition and regularity. If regularity is a critical feature, defining
the boundary between perceptual categories, memory for form of
presentation will be superior when the experimental conditions cross
this boundary-with irregular derivations and base words, for
instance-than when they do not. Similarly, if the concept of a
perceptual record is applicable, the relationship between the translation
classes should be repeated in the second experiment. Specifically, there
should be a negative correlation between RP and memory for form of
presentation.

Method
Subjects
Eight Spanish-English bilinguals, four males and four females, participated in
the experiment. Their mean age was 26.3 years (standard deviation= 5.9). Four
of the subjects-two males and two females-were native Latin Americans.
Three subjects-two males and one female-were native Spaniards. One subject
was Australian-born, but this subject had lived in Latin America for the
preceding seven years. All subjects claimed fluency and at least five years’
experience in English and Spanish.

Stimuli
The stimuli for this experiment were selected from the word set used in
Experiment I. They consisted of the first 24 words and their Spanish equivalents
from the second, third, fourth, and fifth of the five translation classes described
above. Each word was typed in bold, black capital letters in the centre of a
125- x 125-mm white card. Two word arrangements were used, and the items
Bilingual Lexical Representation 383

were distributed so that, across subjects, each item was presented in Spanish and
in English on four occasions. The words were allocated to each sub-set so that
mean frequency of occurrence was the same in each case.

Procedure
The experiment was conducted in two phases in a single session. The phases
involved an incidental learning task and an explicit memory test, respectively.
In the first phase, each subject was presented with a set of 96 cards, 12 cards
from each of the eight conditions given by the factorial combination of language
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

(2) and translation class (4). The words were presented in a new random order
for each subject, and the subjects were instructed to read the cards silently at
their own pace. The subjects were naive with respect to the purpose of the
experiment.
Each subject was also presented with 96 cards in the second phase of the
experiment.These test words were organized so that there were 48 per language,
and so that one half of the words in each language were presented in the same
language on the two occasions, whereas the other half were presented in different
languages on the two occasions. The subjects were instructed to classify the
words into “same” and “different” sets; “same” for words presented in the same
language on each occasion, “different” for words presented in different
languages on the two occasions.
The first and second phases of the experiment each lasted approximately 10
min, and they were separated by a 10-min interval during which
autobiographical details were recorded.

Results
Table I11 summarizes the results from the second experiment. In terms
of the main effect of translation class, the results were as expected.
Memory for language was reliable in the non-cognate conditions, but
barely above chance in the cognate conditions.
An analysis of variance (translation class x test language x repetition
status) showed that the main effect of translation class F(3,21)= 12.61,
M S E = 230,p < 0.01,and the interaction between translation class and
repetition status, F(3, 21)= 6.39, M S E = 489, p < 0.001, were both
significant. Newman-Keuls tests were conducted to compare the means
of the four levels of translation class at each level of repetition status.
There was a significant difference in accuracy between the non-cognate
class and the three cognate classes under “different” conditions, but not
under “same” conditions. There were no other significant main effects
or interactions.
T h e results from the second experiment conformed closely to
qualitative and quantitative predictions based on the hypothesis that
morphology rather than language governs lexical representation. Where
morphologically unrelated translations are concerned, the overall result
(averaged over “same” and “differenty’ treatments) was similar to the
EP A SB/S-C
384 P.Cristoffanini et al.

Table I11
Experiment 2: Memory for language. Accuracy (percent correct) as a
junction of translation class and repetition treatment
Alternative Same
Translation Class Form Form Mean
2 Cognates with regular suffix: 54.2 61.5 57.9
cionltion
3 Cognates with regular suffix:
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

57.3 68.8 63.1


dadlty
4 Cognates with irregular suffixes 66.7 59.4 63.1
5 Morphologically unrelated 95.8 63.6 79.7
translations

summary mean from the published research results. Where the cognates
were concerned, the only intra-lingual basis for prediction yielded a
range from 56.0 to 64.0% correct (Downie et al., 1985), and this range
actually encompasses the means from all three cognate conditions.

GENERAL DISCUSSION
The Comparative Status of Language and Morphology
in Cognate Processing
The results of the experiments demonstrate that cognates should be
classified with inflections and derivations rather than other,
morphologically unrelated translations. In two experiments, using
different test procedures, cognate performance conformed to qualitative
and quantitative predictions based on known patterns of performance
for inflections and derivations. Using cognates that involved no change
in their stem forms and both regular and irregular suffixation rules, the
following results were obtained. First, in all cases the observed RP
values were between those previously reported for (a) regular inflections
in which the stem is the same as the base word (see Table I), and (b)
derivations which include a small change in the stem (see Downie et al.,
1985). Second, the level of reliability obtained in the explicit memory
test was similar to that reported previously for the contrast between base
words and suffixed forms. Memory for stimulus class is unreliable in
Bilingual Lexical Representation 385

each case. People find it difficult to remember whether they saw a base
form or its derivation, in the monolingual example, and a Spanish word
or its English cognate, in the bilingual example. In inferential terms, the
results are inconsistent with the hypothesis that language is a criterion
governing cognate representation.
The conclusion that cognates should be classified with derivations
rather than translations may be challenged on three grounds, and we will
consider each of these in turn. The first challenge stems from the
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

possibility that the observed effects simply reflect orthographic


similarity, that is, similarity in a sub-lexical as distinct from lexical
constituent. This challenge may be set aside for two reasons, however.
First, while priming has been observed at the graphemic level, the effect
is transient (Evett and Humphreys, 1981), and does not extend to
remote repetition intervals (Murrell and Morton, 1974). Second, RP
values similar to those obtained for cognates in Experiment 1 have been
obtained under conditions where orthographic similarity can be
discounted entirely. In a recent study Brown, Sharma and Kirsner
(1984) examined repetition transfer between Hindi and Urdu, languages
that have many common spoken forms but totally different scripts-
Hindi being written from left-to-right in the Devanagari script, whereas
Urdu is written from right-to-left in a variant of Arabic. They did not
discuss their data in relative priming terms, but the mean RP value
based on their summary data is 0.78, a value that compares favourably
with those reported here. In summary, then, orthographic similarity
does not provide an explanation that is either suficient or necessary.
Visually similar but morphologically unrelated words do not yield high
(or even non-zero) RP values, and morphologically related but visually
dissimiliar words can yield RP values comparable to those observed
here.
The second challenge stems from the possibility that the observed
effects are an artifact of the stimulus conditions used during the first
phase of each experiment. In brief, the argument is that subjects were
guided toward a strategy involving morphological decomposition by the
high proportion of cognates included in the study. The hypothesis
cannot be rejected on logical grounds. However, in subsequent research
where we have used (a) sentences during the encoding stage, thereby
reducing the proportion of cognates to less than 5% of all words, and (b)
Italian words, where it is possible to compare cognates that involve
“early”, “middle”, and “late” changes in orthographic form (e.g.,
ELICOTTERO-HELICOPTER), the outcome was similar. The RP
value for cognate pairs that differ in all three word parts was 0.61.
The third challenge concerns the possibility that the results
386 P. Cristoffanini et al.

summarized here can be explained in terms of language-specific lexical


entries, even for cognates. The crucial assumption in this account is that
people do not adhere rigidly to experimenter-defined language, during
either experimentation or normal practice. For example, if it is assumed
that bilinguals have language-specific lexical systems, and that they
normally consult both systems, the frequency effect can be explained
directly, without invoking language-independent processes. The
repetition priming and memory for stimulus class results can be
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

explained in a similar way if it is assumed that subjects either


deliberately or inadvertently consulted the alternative system each time
a word was presented during the study phase of each experiment.
There are three counter-arguments to the above interpretation,
however. The first of these involves an experiment reported by Gerard
( 1983), where she demonstrated that experimenter-defined language
determined the effective frequency level for a selected set of
pseudo-cognates. Using two sets of items (consisting of words that were
high-frequency in Spanish and low-frequency in English, or vice versa,
e.g., FlN), she found that reaction time in a lexical decision task
reflected the frequency status of the words in whichever language was
being tested. The second counter-argument is simply that the language-
independent account is more economical, at least where representation is
concerned. A system that depends on morphology alone will require
fewer entries than a system that includes language-specific entries for
cognates. The third counter-argument is that in its extreme form this
hypothesis is not distinguishable from the hypothesis that lexical
representation in bilinguals is language-independent. Two “units” that
covary in all respects constitute a single unit in functional terms.
In summary, we concede that these problems merit further investigation,
but propose that the morphological account deserves default status.

The Role of Morphology in Word Recognition


Several approaches to the role of morphology in word recognition were
outlined in the introduction. According to one point of view,
decomposition and, possibly, delimitation are crucial concepts. The
lexicon is treated as a dictionary, with abstract and discrete entries, and
each stimulus can be identified only by reference to the appropriate
entry. If this view is correct, it is necessary to identify the critical feature
or features that delimit each perceptual category, with the band-pass
filter providing a useful analogy. The results reported here do not permit
us to reject the decomposition hypothesis in either its lax or its strict
form, or to restrict it to intra-lingual applications. RP approached unity
Bilingual Lexical Representation 387

in each of the four cognate conditions, and in no case is there a significant


differencebetween the relevant intra-lingual and inter-lingual conditions.
There is some suggestion that irregular suffixes are more expensive than
regular suffixes in RP terms (see Table II), but the relevant difference
does not approach significance.
In determining the role of morphology in word recognition, a number
of other factors merit consideration. For example, it could be argued
that morphological priming reflects post-lexical feedback to the target
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

word’s morphological neighbours. Thus, presentation of OBEDIENCIA


during the study phase of an experiment primes not only the
representation for that word, but the representations for OBEDIENT
and OBEDIENCE as well, rather in the manner of spreading activation
in semantic or associative priming. (Although this cannot be the
explanation because of the brevity of the relevant effects: Gough, Alford
and Holley-Wilcox, 1981; Dannenbring and Briand, 1982; Henderson,
Wallis and Knight, 1984). The post-lexical approach certainly provides
a contrast with accounts that describe decomposition in terms of the
retrieval mechanisms associated with a particular task. In Taft and
Forster’s (1975) account, for example, affixes are stripped prior to lexical
access, and, if the pre-lexical assumption is discarded, the economic
gains associated with the model are lost. Each word must be represented
individually. But if the economic problem is set aside, the hypothesis
that morphemic priming reflects post-lexical (and, by implication,
post-lexemic) processes only changes the reference domain. Even if it is
assumed that priming reflects post-lexical processes, the fact that
variation in attributes such as language, modality and morphology
systematically influences RP remains to be explained. Assuming, for
example, that our re-analysis of Stanners et al.3 (1979) results is valid,
why do variations in the relationship between the stem of an inflection or
derivation and the base word influence repetition priming? These
questions can, of course, be addressed in terms of post-lexical processes,
but the answer must involve reference to the set of stimulus
characteristics invoked to explain repetition priming in pre-lexical
terms. No doubt it will be possible to explain the effects in post-lexical
terms, but it will be more expensive.

The Perceptual Record, Repetition Priming,


and Memory for Stimulus Class
As an alternative to the dictionary model, we set out the proposition that
stimulus identification establishes a perceptual record, that that
perceptual record may be expressed in procedural or declarative terms,
388 P. Cristoffanini et al.

and that it may influence performance in a variety of ways, depending on


the way in which the processes invoked by the test task use the record.
Thus, performance in an identification task will be facilitated to the
extent to which identification of a subsequent stimulus involves access to
the same set of processes (and, by implication, the same record), whereas
memory for stimulus class will be impaired by increasing intersection
between the two sets of processes.
In the present study, our claim that the perceptual record influences
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

both repetition priming and memory for stimulus class can be tested
directly by examining the relevant values for each cognate class. The
results of this analysis are shown in Figure 1, where it is apparent that
there is a relationship between the two variables ( r = - 0.99, a!!= 2,
p C 0.05). For comparison, the figure includes a comparable analysis of
the results of two experiments reported by Downie et al. (1985). The
independent variable in that study involved manipulation of phonemic
and stress differences between the stems of English derivations and their
base words, and repetition priming was tested in a word identification
task. However, the studies are similar with regard to number of items,
the interval between the study and test phases, and instructions. The
results are qualitatively comparable to those reported here ( r = - 0.92,
’‘ 2, p < 0.05, 1-tail).
a=
These correlations should be interpreted cautiously. They
demonstrate that there is a relationship between the tasks, but because
they follow averaging across subjects, items, and, possibly, processes,
the correlations do not indicate the strength of that relationship at the
intra-individual level, where the relevant processes must be modelled
(Brown and Kirsner, 1980). On the other hand, the correlation provides
a measure of protection against circularity in our argument. If our
analysis were based on unidimensional effects, involving repetition
priming, for example, our argument would be vulnerable to the type of
criticism that has been levelled against Craik and Lockhart’s (1972)
“depth of processing” hypothesis. That is, that it lacks independent
specification of depth (Baddeley, 1978). But, by demonstrating that the
relationship between the cognate classes occurs under explicit as well as
implicit memory conditions, our claim that performance reflects process
intersection receives independent verification.
The relationship between relative priming and memory for stimulus
class has a number of implications. The first of these concerns its locus,
where they may be considered with reference to the distinction between
process and representation. If it is assumed that there is no overlap
between the processes invoked during test phases of the lexical decision
and memory for stimulus class tasks, it follows that the observed
Bilingual Lexical Representation 389

.o= 100

Cognates and Translations


Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

L L

-
0

L
>
0
,"
Suffixed Derivations

s
0
r = .92
50-1

0.0 0.2 0.4 0.6 0.8 1 .o

Relative Priming

Figure 1. The relationship between memory for form of presentation and


relative priming for Experiments 1 and 2 (refer to Table I1 for translation class)
and for Experiments 1 and 2 in Downie, Milech and Kirsner (1985) where 6 = no
phonological change in the stem, 7 =one phoneme change in the stem, 8 =two
phoneme changes in the stem, and 9 =one phoneme and one stress change in the
stem.

correlation reflects access to either the same representation, or to


equivalent representations. Indeed, even if it is assumed that there is
overlap between some of the early processes invoked in each task
(stimulus identification being the obvious candidate), it is difficult to
imagine how this could lead to the negative relationship actually
observed. Thus, the correlation is consistent with the view that the
pattern is influenced by the nature of the representation formed during
the study phase of the experiment. The precise way in which that
representation influences performance will, of course, depend on the
nature of the processes that operate on it. Where the task involves
procedural recapitulation, as in word identification and lexical decision,
performance will be facilitated because the observer is in effect
performing the same operations on the same stimulus or on a similar
stimulus. But when the observer is required to review the original
record, when memory for stimulus class is tested, performance will be
390 P. Cristoffanini et al.

impaired by increasing intersection between the sets of processes


invoked on the two occasions. Assuming, then, that there is no overlap
between the critical processes invoked during the test phases of the two
tasks, it follows that the correlation is determined by their reliance on a
common representation.
The precise nature of the representation is not revealed by the present
research; however we favour the view that it consists of a record of the
processes invoked during initial stimulus identification. This position
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

has a number of implications. First, it means that a wide range of


stimulus qualities will be preserved in the record, although only in terms
of the processes that were invoked to establish identity. Secondly, it may
be reasonable to assume that each record consists of a very large set of
processes; one implication of this- being that although each change
between study and test form will have a cost, in RP terms, many of these
will not yield statistically reliable differences in performance because the
variability associated with the dependent measures is so great relative to
the effect of the change on performance. Third, a change in, for example,
modality, an attribute that subsumes a range of dependent properties
(e.g., size, colour, position and font for print; fundamental pitch,
position, accent and stress for speech), can be expected to yield
substantial and significant RP effects simply because so many of the
relevant processes are not consulted when the alternative form is
submitted for identification. Fourth, a given RP value may be achieved
in a variety of ways; thus an observation that changes in, say, modality
and lexical form both yield an RP value of 0.50 does not mean that the
same subsets of processes are involved in the two cases. They may be,
but the only inference that is logically demanded is that the same
proportion of processes is common to the study and test solutions in each
case.
The third major implication of the observed relationship between
relative priming and memory for stimulus class is in the support it
provides for the view that perceptual categorization involves access to an
instance-based memory system, a position espoused by Jacoby and
Brooks (1984), although for rather different reasons from those
developed here.
In summary, the results are consistent with the proposition that
morphology rather than language governs lexical function. However,
they also suggest that a strictly analytic approach, involving abstract and
stable recognition units, may be inappropriate. A non-analytic approach
in which more or less similar perceptual records are reviewed provides a
plausible alternative and receives support from the observed
co-variation between the relative priming and memory for stimulus
class.
Bilingual Lexical Representation 391

REFERENCES

Albert, M. L. and Obler, L. K. (1978).The bilingual brain. New York Academic


Press.
Anderson, J. R. (1978). Arguments concerning representations for mental
imagery. Psychological Review, 85(4), 249-277.
Baddeley, A. D. (1978).The trouble with levels: A reexamination of Craik and
Lockhart’s framework for memory research. Psychological Review, 85(3),
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

139-152.
Barnett, G.A. (1977). Bilingual semantic organisation: A multi-dimensional
analysis. Journal of Cross-Cultural Psychology, 8(3), 315-327.
Bradley, D. (1979). Lexical representation of derivational relations. In M.
Aranoff and M. L. Kean (Eds.) Juncture. Cambridge, Mass: M.I.T. Press.
Brown, H. L. and Kirsner, K. (1980).A within subjects analysis of the
relationship between memory span and processing rate in short-term
memory. Cognitive Psychology, 12, 177-187.
Brown, H. L.,Sharma, N. K. and Kirsner, K. (1984).The role of script and
phonology in lexical representation. Quarterly Journal of Experimental
Psychology, 36A, 491-505.
Caramazza, A. and Brones, I. (1979).Lexical access in bilinguals. Bulletin of the
Psychonomic Society, 13,2 12-2 14.
Craik, F. I. M. and Kirsner, K. (1974).The effect of speaker’s voice on word
recognition. Quarterly Journal of Experimental Psychology, 26, 274-284.
Craik, F. I. M. and Lockhart, R. S. (1972).Levels of processing: A framework
for memory research. Journal of Verbal Learning and Verbal Behavior, 11,
671-684.
Dannenbring, G.L. and Briand, K. (1982). Semantic priming and the word
repetition effect in a lexical decision task. CanadianJournal of Psychology, 36,
435444.
Downie, R., Milech, D. and Kirsner, K. (1985).Unit definition in the mental
lexicon. Australian Journal of Psychology, 37(2), 141-155.
Evett, L. J., and Humphreys, G. W. (1981).The use of abstract graphemic
information in lexical access. Quarterly Journal of Experimental Psychology,
33A, 325-350.
Fisher, R. P. and Cuervo, A. (1983).Memory for physical features of discourse
as a function of their relevance.Journal of Experimental Psychology: Learning,
Memory and Cognition, 9(1), 130-138.
Gerard, L. D. (1983).Bilingual word recognition: Lexical access. Paper
presented to the April meeting of the Eastern Psychological Association,
Philadelphia.
Gough, P. B., Alford, J. A. and Holley-Wilcox, P. (1981).Words and contexts.
In 0. J. L. Tzeng and H. Singer (Eds.), Perception of print. Hillsdale, N. J.:
Lawrence Erlbaum Associates.
Henderson, L. (1984).Writing systems and reading processes. In L. Henderson
(Ed.), Orthographies and reading. Hillsdale, N. J.: Lawrence Erlbaum
Associates.
392 P. Cristoffanini et al.

Henderson, L. (1985). Toward a psychology of morphemes. In A. W. Ellis


(Ed.), Progress in the psychology of language. Vol. 1. Hillsdale, N. J.: Lawrence
Erlbaum Associates.
Henderson, L., Wallis, J. & Knight, D. (1984). Morphemic structure and lexical
access. In H. Bouma and D. Bouwhuis (Eds.), Attention and performance X .
London: Lawrence Erlbaum Associates.
Jackson, A. and Morton, J. (1984). Facilitation in auditory word recognition.
Memory & Cognition, 12(6), 568-574.
Jacoby, L. L. and Brooks, L. R. (1984). Nonanalytic cognition: Memory,
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

perception and concept learning. In G. H. Bower (Ed.), The psychology of


learning and motivation: Advances in research and theory, Vol. 18. New York:
Academic Press.
Juilland, A. and Chang-Rodriguez, E. (1964). The Romance languages and their
structures. Directed by A. Juilland, first series S1: Frequency dictionary of
Spanish words. The Hague: Mouton.
Kempley, S. T. and Morton, J. (1982). The effects of priming with regularly and
irregularly related words in auditory word recognition. British Journal of
Psychology, 73, 441454.
Kintsch, W. (1970). Recognition memory in bilingual subjects.Journal of Verbal
Learning and Verbal Behavior, 9, 405409.
Kirsner, K. (1973). An analysis of the visual component in recognition memory
for verbal stimuli. Memory Q Cognition, 1,449-453.
Kirsner, K., Brown, H. L., Abrol, S., Chaddha, N. N. and Sharma, N. K.
(1980). Bilingualism and lexical representation. Quarterly Journal of
Experimental Psychology, 32, 585-594.
Kirsner, K. and Dunn, J. C. (1985). The perceptual record: A common factor in
repetition priming and attribute retention. In M. I. Posner and 0. S. M.
Marin (Eds.), Mechanisms of attention: Attention and performance X I .
Hillsdale, N. J.: Lawrence Erlbaum Associates.
Kirsner, K., Smith, M. C., Lockhart, R. S., King, M-L and Jain, M. (1984).
The bilingual lexicon: Language-specific units in an integrated network.
Journal of Verbal Learning and Verbal Behavior. 23, 519-539.
Kolers, P. A. (1978). On the representations of experience. In D. Gerver and H.
W. Sinaiko (Eds.), Language interpretation and communication. New York:
Plenum.
Kolers, P. A. (1979). Reading and knowing. Canadian Journal of Psychology,
33(2), 106-1 17.
Kolers, P. A. and Roediger, H. L. (1984). Procedures of mind. Journal of Verbal
Learning and Verbal Behavior, 23,425449.
Kucera, H., and Francis, W. N. (1967). Computational analysis of present-day
American English. Providence, R. I.: Brown University Press.
McCormack, P. D. (1976). Language as an attribute of memory. Canadian
Journal of Psychology, 30(4), 238-248.
McCormack, P. D. (1977). Bilingual linguistic memory: The independence-
interdependence issue revisited. In P. Hornby (Ed.). Bilingualism:
Psychological, social and educational implications. New York. Academic Press.
Bilingual Lexical Representation 393

Meyer, D. E. and Ruddy, M. E. (1974). Bilingual word recognition:


Organization and retrieval of alternative lexical codes. Paper presented to
April meeting of the Eastern Psychological Association, Philadelphia.
Monsell, S. (1985). Repetition and the lexicon. In A. W. Ellis (Ed.), Progress in
the psychology of language. Vol. 2. London: Lawrence Erlbaum Associates.
Murrell, G. A. and Morton, J. (1974). Word recognition and morphemic
structure. Journal of Experimental Psychology, 102, 963-968.
Perecman, E. (1984). Spontaneous translation and language mixing in
polyglotaphasia. Brain and Language, 23, 43-63.
Downloaded By: [Universidad Autonoma del Estado de Mexico] At: 17:45 8 October 2010

Posner, M. (1969). Abstraction and the process of recognition. In K. W. Spence


and J. T. Spence (Eds.), The psychology of learning and motivation, Vol. 3.
New York: Academic Press.
Rose, R. G. and Carroll, J. F. (1974). Free recall of a mixed language list.
Bulletin of the Psychonomic Society, 3(4), 267-268.
Saegert, J., Hamayan, E. and Ahmar, H. (1975). Memory for language of input
in polyglots. Journal of Experimental Psychology: Human Learning and
Memory, 1(5), 607-613.
Scarborough, D. L., Cortese, C. and Scarborough, H. S. (1977). Frequency and
repetition effects in lexical memory. Journal of Experimental Psychology :
Human Perception and Performance, 3, 1-17.
Scarborough, D. L., Gerard, L. and Cortese, C. (1984). Independence of lexical
access in bilingual word recognition. Journal of Verbal Learning and Verbal
Behavior, 23, 84-99.
Smith, C., Bermejo, M. M., and Chang-Rodriguez, E. (1971). Collins
Spanish-English English-Spanish dictionary. London: Collins.
Stanners, R. F., Neiser, J. J., Hernon, W. P. and Hall, R. (1979). Memory
representation for morphologically related words. Journal of Verbal Learning
and Verbal Behavior, 18, 399412.
Taft, M. and Forster, K. I. Lexical storage and retrieval of prefixed words.
Journal of Verbal Learning and Verbal Behavior, 14,638-647.
Tulving, E. (1983). Elements of episodic memory. Oxford: Clarendon Press.
Underwood, B. J. (1969). Attributes of memory. Psychological Review, 76,
559-573.
Winograd, E., Cohen, C. and Barresi, J. (1976). Memory for concrete and
abstract words in bilingual speakers. Memory 0 Cognition, 4(3), 323-329.

Revised manuscript received 6 December 1985

S-ar putea să vă placă și