Professional Documents
Culture Documents
Colombo Et Al - 2003 - Two Cycle
Colombo Et Al - 2003 - Two Cycle
Colombo Et Al - 2003 - Two Cycle
a
Dipartimento di Psicologia Generale, Università di Padova, Italy
b
Dipartimento di Psicologia Generale, Università di Padova, Italy
and Institute of Cognitive Neuroscience, University College London,
UK
c
Istituto di Psicologia, Università di Urbino, Italy
Published online: 10 Sep 2010.
To cite this article: Lucia Colombo , Marco Zorzi , Roberto Cubelli & Cristina Brivio (2003) The
status of consonants and vowels in phonological assembly: Testing the two-cycles model with Italian,
European Journal of Cognitive Psychology, 15:3, 405-433, DOI: 10.1080/09541440303605
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
EUROPEAN JOURNAL OF COGNITIVE PSYCHOLOGY, 2003, 15 (3), 405±433
Marco Zorzi
Dipartimento di Psicologia Generale, UniversitaÁ di Padova, Italy and
Institute of Cognitive Neuroscience, University College London, UK
Roberto Cubelli
Istituto di Psicologia, UniversitaÁ di Urbino, Italy
Cristina Brivio
Dipartimento di Psicologia Generale, UniversitaÁ di Padova, Italy
Recent studies by Berent and Perfetti (1995) and Lee, Rayner, and Pollatsek (2001)
have shown a processing priority of consonants over vowels in tasks like backward
masking and reading, respectively. One possible explanation for this advantage of
consonants is in terms of the specific characteristics of the English spelling (i.e., a
great inconsistency in the spelling±sound relations for vowels), and therefore is not
predicted for a language, like Italian, with a transparent orthography. The aim of
the present experiments was to investigate the consonant advantage in Italian with
the backward masking paradigm. The first two experiments, with a perceptual
identification and a naming task, showed a masking reduction effect for masks
sharing vowels with the target at the shortest SOAs, in addition to a pseudo-
homophone advantage as compared to the control condition. In the third experi-
ment, where a lexical decision task was used, the advantage of the mask sharing
vowels disappeared. The fourth experiment was a replication of Experiment 1 with
different materials. Overall, the results suggest that the advantage of consonants
can only be found in English, whereas in Italian, for which the computation of
phonology of vowels and consonants is similar, no effect is apparent. Thus, the
processing priority for consonants seems to depend on the inconsistencies of vowel
phonology in English.
One highly debated issue concerning the reading of words regards the nature of
the sublexical mechanism, i.e., the mechanism that is assumed to compute and
assemble the phonemes of a word from the corresponding graphemes. In the dual
route model of reading aloud, phonological assembly is carried out through
grapheme to phoneme correspondence (GPC) rules, that is, a system based upon
explicit rules specifying the dominant (e.g., most frequent) relationships between
letters and sounds (Coltheart, 1978; Coltheart, Curtis, Atkins, & Haller, 1993;
Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001). According to one view, the
GPC route is slow and serial (i.e., it delivers one phoneme at a time; Rastle &
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
Besner, 1987; Carr & Pollatsek, 1985; Treiman, Mullenix, Bijeljac-Babic, &
Richmond-Welty, 1995). Therefore, the phonological representation for con-
sonants would be computed in the first cycle, whereas for vowels it would be
computed by a slow, controlled process, in the second cycle.
Berent and Perfetti (1995) found evidence in favour of their model in a
number of experiments with the backward masking paradigm, in which the
phonological overlap between target and mask was manipulated. The main
finding was that the consonant-preserving nonword mask (RIIK) yielded higher
identification accuracy of the target word (rake) than the vowel-preserving mask
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
(RAIB) at brief exposure durations (15 and 30 ms for the target, 30 ms for the
mask). In contrast, at the longer exposure durations (45 ms for the target, 30 and
60 ms for the mask), the advantage of the consonant-preserving mask dis-
appeared. These results have important implications both at a general level (i.e.,
regarding the nonlinear, and separate phonological representations of consonants
and vowels) and for models of reading, because they challenge the serial, left-to-
right processing assumption of the dual route model (Rastle & Coltheart, 1999).
The series of experiments reported below was designed to investigate the
consonant advantage with a different orthographic system, the Italian system. The
most interesting feature of the Italian writing system in the present context is that
the spelling±sound correspondences are equally reliable for both consonants and
vowels. In most cases, there is a one-to-one correspondence between letters and
phonemes; if anything, vowels are even more unambiguous than consonants,
because the pronunciation of some consonants (e.g., C, G) is context sensitive.
Therefore, we sought to investigate the status of consonants and vowels in
phonological assembly, in the absence of the computational constraints that are
produced by the peculiarities of the spelling±sound mapping in English. If the
consonant advantage is reliable, and if it is due to the inconsistency of the English
spelling±sound correspondences for vowels, no processing priority in assembling
the phonology of consonants compared to that of vowels should be found in Italian.
As an alternative, the consonant advantage might be a reflection of processes
that are not language-specific. A striking neuropsychological dissociation
between consonants and vowels has been observed in the spelling errors pro-
duced by Italian-speaking patients (Cotelli, Abutalebi, Zorzi, & Cappa, 2003;
Cubelli, 1991). The patients showed selective vowel deletion (case CF: Cubelli,
1991) or substitution (case CW: Cubelli, 1991; case LiB: Cotelli et al., 2003).
Further, the patient described by Cotelli et al. showed a similar, although milder,
error pattern in reading. Given that the converse dissociation has never been
reported, the spared processing of consonants in Italian patients seems to
somewhat mirror the faster and more automatic processing of consonants
described in the backward masking studies on skilled English readers. Further,
evidence for a representational distinction of consonants and vowels has also
been found in the spelling of children (Colombo, Cubelli, & Inzaina, 1998). It
has been argued (Colombo, 2000) that this representational distinction might
reflect a corresponding processing difference in the computation of phonology.
408 COLOMBO ET AL.
EXPERIMENTS 1 AND 2
The aim of Experiment 1 was to replicate as closely as possible Berent and
Perfetti's (1995) Experiment 1 in Italian, with the backward masking paradigm.
Participants were required to write down what they perceived of the written
stimulus. In Experiment 2 the same materials and conditions of presentation
were maintained, but participants had to read aloud the target word. In the latter
task, the phonological representation of the word is necessarily required by the
task itself. If the assumption that a phonological code is rapidly activated from
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
Method
Participants. One-hundred-and-three students of the University of Padua
volunteered to participate in Experiment 1. Seventy-two students of the Uni-
versity of Padua participated in Experiment 2. All were native Italian speakers
and had normal or corrected-to-normal vision. Participants of Experiment 1
were randomly assigned to the four different conditions of target±mask exposure
duration: 22 to the 16:33 ms condition, 22 to the 33:33 ms condition, 28 to the
50:33 ms condition, 31 to the 50:67 ms condition.1 Participants of Experiment 2
were randomly assigned to three different conditions of target±mask exposure
duration: 32 to the 16:33 ms condition, 20 to the 33:33 ms condition, and 20 to
the 50:33 ms condition.2
Material. Fifty-six target words were used in the experiment, all nouns and
adjectives with ``regular'' stress (i.e., with stress on the penultimate syllable; see
Colombo, 1992), with frequencies between 4 and 6 in a corpus of 496,335
entries (De Mauro, Mancini, Vedovelli, & Voghera, 1993). There were two
levels of word length (two- or three-syllable words) and two levels of word
structure (simple: CVCV and CVCVCV, or complex). Simple and complex
structure refers to the absence or presence of one or more consonant clusters,
1
The number of participants in each exposure condition varies because initially we thought to
maintain a cutoff of accuracy between 10% and 90%, calculated separately for each condition and
participant. At the longer durations the accuracy of many participants was above 90% in some
conditions, therefore according to this criterion the data for these participants should have been
removed, and so more participants were tested. However, as in Berent and Perfetti (1995) the cutoff
is calculated through all conditions, and we wanted to make a replication as close as possible, it was
decided that all the available data were appropriate and could be used in the analyses.
2
The SOA of 50.33 was used by mistake, rather than the longer 50.67 which gave more robust
effects in Experiment 1.
CONSONANTS AND VOWELS IN ITALIAN 409
3
As the characteristics of the Italian writing system (i.e., the absence of multiple readings of the
graphemes) do not allow a straightforward generation of homophonic nonwords, we constructed a set
of pseudohomophones by changing the Italian graphemes ``c'', ``i'', and ``v'' with the graphemes
``k'', ``y'', and ``w'', respectively. For example, from the real Italian words pesca (peach), ansie
(anxieties) and corvo (craw) we derived the nonwords PESKA, ANSYE, and CORWO, that are
pronounced in the same way. Although the latter graphemes do not belong to the Italian alphabet,
they are frequently encountered by Italian readers in words borrowed from foreign languages (e.g.,
English, German) and are usually pronounced by Italian speakers like the Italian phonemes /k/, /i/,
and /v/. In a pre-experiment we tested the naming latencies and accuracy of pseudohomophones
(peska) mixed with ``legal'' nonwords (desca, i.e., containing only the graphemes of the Italian
alphabet), and with ``strange'' nonwords (deska, i.e., containing also foreign graphemes) with 30
students of the University of Padua. The mean latencies and error percentages (in parentheses) for the
pseudohomophones, legal nonwords, and ``strange'' nonwords were, respectively, 628 ms (3.86),
647 ms (6.61), and 670 ms (9.51). Planned comparisons revealed that the pseudohomophone stimuli
yielded faster latencies and higher accuracy than the ``strange'' nonwords, t(29) = 2.2, SE = 18.8, p <
.05, for latencies; t(29) = 3.8, SE =1.5, p < .001, for accuracy, while the difference in accuracy
between pseudohomophones and the ``standard'' nonwords approached significance, t(29) = 1.8, SE
= 1.5, p = .08. The results showed an advantage of the pseudohomophones only compared to the
strange nonwords, which suggests, on one hand, a disadvantage at the level of orthographic pro-
cessing due to the presence of foreign letters, and on the other hand an advantage at the phonological
level because the phonological representation of the pseudohomophone corresponded to that of a real
word.
410 COLOMBO ET AL.
thus preserved both consonant and vowel information at the phonological level
(e.g., ``KAPRA'' and ``YNSETTO'' for ``capra'' and ``insetto'', respectively)
by the inclusion of foreign letters, although it was not graphemically equated to
the control mask due to the regularity of Italian.
(4) Control mask: The mask shared the initial phoneme with the target
word, and sometimes also other graphemes, because of phonotactic constraints
that had to be respected in order for the mask to be a legal nonword. The number
of letters and the position of vowels and consonants in the control string were the
same as the target's (e.g., ``RABRO'' was the mask for ``ritmo'').
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
masks in blank trials were constructed using stimuli that were not included in the
experimental lists, by changing the consonant or the vowels of words, or making
a pseudohomophone. Their purpose was to ensure that participants did not guess
the target on the basis of the mask (Perfetti & Bell, 1991). Another list of 30
mask and target pairs was constructed to be used as practice trials. The masks
were unrelated to the targets in the practice trials.
present paper. Reaction times were recorded by the computer via a microphone
connected to a voice key.
Procedure. Type of mask, word length, and word structure were within-
participants variables. Exposure duration of the target±mask pair was a between-
participant variable: For the brief durations, the target was exposed either for 16
ms (1 refresh cycle) or 33 ms and the mask was exposed for 33 ms, giving two
combinations of target±mask duration (16:33 and 33:33). For the long durations,
the target was exposed for 50 ms and the mask was exposed either for 33 ms or
67 ms, giving two combinations of target±mask duration (50:33 and 50:67).
Participants were seated in front of a 14-inch display, at a distance of about
50 cm. Each trial started with the presentation of four dots at the corners of a
rectangular area located at the centre of the computer screen, signalling the
location of stimulus presentation. The target stimulus subtended a visual angle
of about 1.38 to 2.28, depending on the length of the words. The word was
presented in the centre of the screen (in lowercase letters), followed immediately
(ISI = 0 ms) by a pseudoword mask (in uppercase letters), then replaced by a
pattern mask consisting of a row of Xs, which indicated the end of the trial and
remained on the screen until the participant made a response and the experi-
menter pressed a button to start the next trial. Three short tones before pre-
sentation of the target word alerted the participant that the next trial had started.
After each trial the sheet was removed, to avoid any interference with the
following trial.
The experimental session began with 30 practice trials, divided into five
blocks, during which target duration was gradually reduced until it reached the
value assigned to the participant. Participants of Experiment 1 were informed of
the sequence of events and asked to write down on a sheet of paper whatever they
could read, including partial responses if they could not report seeing a word.
In Experiment 2, the target was presented for 16, 33, or 50 ms, and the mask
was presented for 33 ms, yielding three target±mask duration conditions (16:33,
33:33, and 50:33). Participants were instructed to pronounce the target words as
quickly and accurately as possible, and naming latencies were recorded by the
computer through a voice-key. The experimenter sat next to each participant
throughout the experiment and noted the errors and the accidental triggering of
412 COLOMBO ET AL.
the voice key. The procedure was in all other respects identical to that for
Experiment 1.
Results
Experiment 1: Perceptual identification task. Only the accuracy data for
each participant whose percentage of correct responses was between 10% and
90% are reported (see Table 1). This cutoff was the same as that used by Berent
and Perfetti (1995) in order to eliminate floor and ceiling performance. Two
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
participants in the 16:33, in the 33:33, and in the 50:67 ms condition, and three
participants in the 50:33 ms condition were excluded from the analyses due to
the accuracy cutoff. The assignment of participants to the lists were still nearly
balanced after the cutoff. The data were submitted to a repeated measures
ANOVA, with mask type (four levels: consonant-preserving, vowel-preserving,
pseudohomophone, and control), word structure (two levels: simple and com-
plex), and word length (two levels: two-syllable and three-syllable) as in-
dependent variables. As noted above, target and mask were presented in four
different duration combinations, that is, 16:33 ms, 33:33 ms, 50:33 ms, and
50:67 ms. The values of F1 and F2 indicate the analysis by participants and by
items, respectively.
The overall analysis included mask type as a within-participants variable and
SOAs as a between-participants variable. The ANOVA indicated main effects of
mask type, F1(3, 282) = 39.3, MSE = 664.2, p < .001; F2(3, 156) = 34.3, MSE =
388.4, p < .001, and of duration, F1(3, 94) = 7.4, MSE = 6497.1, p < .001;
F2(3, 156) = 87.2, MSE = 296.3, p < .001. The main effects of structure and
length of words were significant by participants only, F(1, 94) = 23.1, MSE =
825.7, p < .001 for structure; for length the effect was marginal, F(1, 94) = 3.73,
MSE = 651.3, p = .057. The critical interaction between duration and mask was
significant, F1(9, 282) = 3.6, MSE = 664.2, p < .001; F2(9, 468) = 2.9, MSE =
431.9, p <.01. Also significant by participants, and marginally by items, was the
interaction structure by length, F1(1, 94) = 25.9, MSE = 813.5, p < .001;
F2(1, 52) = 3.2, MSE = 3011.3, p < .1.
TABLE 1
Accuracy percentage in the four mask conditions in the perceptual
identification task (Experiment 1)
Four separate analyses were run on the accuracy for the short and the long
durations, with mask type, length, and complexity as within-participant vari-
ables. As the effect of word structure/length is not theoretically relevant here,
further reports of analyses of these factors will be omitted. For the 16:33 ms
duration, the analysis showed a reliable effect of mask type, F1(3, 63) = 10.7,
MSE = 563.3, p < .001; F2(3, 156) = 6.3, MSE = 509.3, p < .001.
Planned comparisons revealed that all comparisons of experimental masks
with the control mask were significant: For the consonant-preserving mask, only
by participants: t1(21) = 2.7, SE = 3.5, p < .05; t2(55) = 1.9, SE = 4.16, .05 < p <
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
.1, for the vowel-preserving mask, t1(21) = 6.8, SE = 2.8, p < .001 and t2(55) =
5.2, SE = 3.7, p < .001, and for the pseudohomophone mask, t1(21) = 5.1, SE =
2.9, p < .001; t2(55) = 3.2, SE = 4.2, p < .005. Also the difference between the
consonant-preserving mask and the vowel-preserving mask was significant,
t1(21) = 3.43, SE = 2.84, p < .01; t2(55) = 2.44, SE = 4.10, p < .05, showing an
unexpected advantage for the vowel-preserving mask. No other difference was
significant.
For the 33:33 ms duration, the analysis showed a significant effect of mask
type, F1(3, 63) = 12.8, MSE = 647.8, p < .001; F2(3, 156) = 9.5, MSE = 494.3, p
< .001. Planned comparisons revealed that all types of masks yielded different
accuracy data compared to the control mask: The difference was reliable for the
consonant-preserving mask, t1(21) = 4.8, SE = 3.4, p < .001, and t2(55) = 3.6, SE
= 4.1, p < .001, for the vowel-preserving mask, t1(21) = 6.3, SE = 3.6, p < .001
and t2(55) = 5.3, SE = 4.2, p < .001, for the pseudohomophone mask, t1(21) =
5.3, SE = 3.9, p < .001 and t2(55) = 4.3, SE = 4.3, p < .001. Furthermore, there
was a significant difference between the consonant-preserving mask and the
vowel-preserving mask in the participant analysis, t1(21) = 2.2, SE = 2.8, p < .05,
t2(55) = 71.22, SE = 4.37, p > .1, showing again an advantage for the vowel-
preserving mask. No other difference was significant.
For the 50:33 ms duration, neither participants nor items analyses yielded any
significant effect. The analyses of the 50:67 ms duration condition, on the other
hand, revealed a significant effect of mask type, F1(3, 84) = 27.8, MSE = 703.9,
p < .001; F2(3, 156) = 28.9, MSE = 306.6, p < .001. Planned comparisons
showed reliable differences between consonant-preserving mask and control
mask, t1(28) = 6.2, SE = 3.6, p <.001; t2(55) = 6.7, SE = 3.7, p < .001, between
vowel-preserving mask and control mask, t1(28) = 6.8, SE = 3.5, p < .001 and
t2(55) = 8.3, SE = 32.9, p < .001, and between pseudohomophone mask and
control mask, t1(28) = 7.2, SE = 3.9, p =.001; t2(55) = 9.5, SE = 3.1, p < .001.
However, the difference between consonant-preserving mask and vowel-pre-
serving mask was not significant.
Experiment 2: Naming task. The accuracy data for each participant were
analysed if the percent of correct responses was between 10% and 90% through
all conditions (to eliminate floor and ceiling performance). The data for five
414 COLOMBO ET AL.
participants with zero accuracy in the control condition of the shortest condition
were removed, leaving twenty-seven participants. The design was the same as in
Experiment 1. The analyses on latencies are also reported, although the main
focus was on accuracy, for a comparison with the perceptual identification task.
The data are shown in Table 2. As noted above, target and mask were
exposed for three different duration combinations, that is, 16:33 ms, 33:33 ms,
and 50:33 ms. An overall ANOVA was run on the accuracy data with mask type,
structure, and length as within-participants variables, and SOA as a between-
participants variable, in the analysis by participants. The critical interaction
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
between mask type and duration was significant, F1(6, 192) = 3.1, MSE = 642.1,
p < .01; F2(6, 312) = 2.74, MSE = 372.0, p < .05. There were significant main
effects of mask type, F1(3, 192) = 14.6, MSE = 642.2, p < .001; F2(3, 156) =
14.9, MSE = 343.4, p < .001, and duration, F1(2, 64) = 82.2, MSE = 2293.8, p <
.001; F2(2, 104) = 202.8, MSE = 519.3, p < .001. The effect of word structure
was significant only in the analysis by participants, F1(1, 64) = 13.6, MSE =
638.9, p < .001. The interaction structure by length was significant in the
analysis by participants, F1(1, 64) = 23.2, MSE = 799.4, p < .001, while it was
marginal by items, F2(1, 52) = 3.5, MSE = 2374.0, p < .1. Last, there was also a
significant interaction between duration, structure, and length, F1(2, 64) = 5.7,
MSE = 799.4, p < .01; F2(2, 104) = 3.7, MSE = 519.3, p < .05. On the latency
data, the mask effect was not significant (both Fs < 1).
For the 16:33 ms duration, the analysis on accuracy data showed a significant
effect of mask type, F1(3, 78) = 17.6, MSE = 756.1, p < .001; F2(3, 156) = 19.3,
MSE = 311.8, p < .001. Planned comparisons showed lower accuracy for the
control mask compared to all the other mask types: For the same consonant
mask, t1(26) = 4.6, SE = 3.9, p < .001; t2(55) = 5.7, SE = 3.2, p < .001, for the
same vowel mask, t1(26) = 7.2, SE = 3.7, p < .001, t2(55) = 6.8, SE = 3.7, p <
.001, and for the pseudohomophone mask, t1(26) = 4.7, SE = 4.0, p < .001, t2(55)
= 5.8, SE = 3.2, p < .001. The difference between the same consonant and the
same vowel masks was significant also in this experiment, and in the same
direction as in Experiment 1, t1(26) = 72.1, SE = 4.2, p < .05 and t2(55) = 71.9,
SE = 3.5, p = .056. Mean latencies for the four masking conditions were 798,
803, 768, and 776, for the consonant-preserving, vowel-preserving, pseudo-
TABLE 2
Accuracy percentage in the three mask conditions in the naming task
(Experiment 2)
between the pseudohomophone and the control masks, t1(19) = 2.8, SE = 3.44, p
< .05, t2(55) = 2.02, SE = 4.78, p < .05. The difference between the same
consonant and the same vowel masks was not significant, nor was any other
comparison. In the analysis of latency mean naming times were 651, 650, 625,
and 656, for the same-consonant, same-vowel, pseudohomophone, and control
conditions, and the mask effect was not significant (both Fs < 1).
For the 50:33 ms duration, the mask effect was not significant on either the
accuracy, F1(3, 57) = 2.31, MSE = 75.8, .05 < p < .1, F2(3, 165) = 1.69, MSE =
290.5, p > .1, or the latency data (both Fs < 1; mean latencies for the same
consonant, same vowel, pseudohomophone, and control conditions were 554,
551, 559, and 548, respectively. A priori comparisons on the accuracy data
showed that the difference between the same consonant and the control mask
was significant, t1(19) = 2.11, SE = 3.21, p < .05; t2(55) = 2.27, SE = 2.98, p <
.05, and the difference between same vowel and control was significant only by
participants, t1(19) = 2.05, SE = 2.43, p = .054; t2(55) = 1.43, SE = 3.5, p > .1.
No other comparison was significant.
Discussion
The response accuracy data of Experiment 1 showed very similar effects of
the independent variables in the two conditions of brief exposure duration
(16:33 and 33:33 ms). First, there was a significant masking reduction effect
for masks that shared features with the target word, as compared to the con-
trol mask. In particular, the consonant-preserving and the vowel-preserving
masks produced higher accuracy compared to the control, but, in contrast
with Berent and Perfetti (1995), when performance to the two types of masks
was compared, a reversal of the pattern of effects was found, showing an
advantage for the vowel-preserving as compared to the consonant-preserving
masks.
The exposure duration of 50:33 ms produced no relevant effect. Therefore,
the identification accuracy of the target word was not modulated by the char-
acteristics of the mask, for this particular interval. A possible interpretation of
this null result is that masking, although interfering with the recognition process,
did not facilitate the reinstatement of the phonological information produced by
416 COLOMBO ET AL.
the longer processing (50 ms) of the target word. The mean identification
accuracy (around 65%) was similar across the four mask types. Berent and
Perfetti (1995) found similar difficulties in interpreting the data from their
condition of intermediate exposure duration.
A reliable effect of masking was found again with the longest exposure
duration (50:67 ms). All types of mask that reinstated phonological information
produced higher accuracy compared to the control mask. In contrast to the brief
exposure durations, however, there was no difference between the vowel-
preserving mask and the consonant-preserving mask. In sum, in Experiment 1
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
al. (2001) pointed out, several aspects of their study can be subject to criticism,
and the generality of their data therefore may be limited.
In contrast, the pattern of results obtained by Berent and Perfetti (1995), with
a processing priority for consonants over vowels, has been replicated for English
in a study by Lee et al. (2001) with a missing letter paradigm in a sentence
context in which the dependent variable was gaze duration. At brief presenta-
tions of a prime in which either consonants (b-lb) or vowels (bu-b) of the target
word (bulb) were preserved, there was an advantage of same consonant over
same vowel primes. In Lee et al.'s study the position of the missing letter was
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
EXPERIMENT 3
The failure of Experiments 1 and 2 to show any evidence of a consonant
advantage might be an indication that the content and the time course of
phonological assembly in reading Italian words is different compared to English.
In both Experiments 1 and 2 we found a masking reduction effect. This effect is
probably located at the phonological level, since the same pattern of data that we
found in the perceptual identification task was apparent in the naming task,
where the phonological representation of the target words must necessarily be
computed. The masking reduction effect was also present in the pseudohomo-
phone mask condition, despite the fact that the graphemic form of pseudo-
homophone masks was not really equated to the other types of masks, due to the
constraints of the Italian language. On the contrary, the pseudohomophone mask
was graphemically less similar to the target than the other types of masks. Thus,
a facilitation for the pseudohomophone condition in the masking experiments,
despite the disadvantage at the graphemic level, implies an even stronger effect
of phonology. The present experiments, therefore, in general do show evidence
in support of the claim of an early emergence of phonology during word
identification.
In Experiment 3 the backward masking paradigm was slightly changed, so
that the task assigned to the participants was lexical decision, rather than
spelling, or naming. The masked target on which lexical decision was made
could be either a word, or a nonword. The aim of the experiment was to see
whether the vowel-preserving versus the consonant-preserving advantage that
appeared in both Experiments 1 and 2 could be replicated, when participants do
not need a well-developed phonological representation as a response, but can
418 COLOMBO ET AL.
Method
Participants. Eighty-four students of the University of Padua volunteered to
participate in the experiment. All had normal or corrected-to-normal vision.
Twenty-two of the participants were assigned to the target±mask exposure
duration condition of 16:33 ms, twenty-two to the 33:33 ms condition, twenty to
the 50:33 ms condition, and twenty to the 50:67 ms condition.
Procedure. The sequence of events for each trial was exactly the same as in
Experiment 1. At the presentation of the pattern mask, participants had to per-
form a lexical decision (i.e., decide whether the target stimulus was a word or a
nonword) by pressing one of two labelled buttons, one on the left side (for
``nonword'') and one on the right side (for ``word'') of the computer keyboard.
Reaction times were recorded to measure the response latencies from pre-
sentation of the target word. To assess the reliability of the responses provided
by the participants, the target word was presented again on the computer screen
(for an unlimited time) after the button-press response. Participants were asked
to say whether the target word was the same they had seen during tachistoscopic
presentation. This second response was provided verbally, and it was checked by
CONSONANTS AND VOWELS IN ITALIAN 419
the experimenter, who had then to press either one of two computer keys,
corresponding to recognition, or failed recognition, in order to include these data
in the analyses. The procedure was in all other respects identical to that for
Experiment 1.
Results
Although both accuracy and latencies were measured, the dependent variable of
interest was accuracy, as the aim of the experiment was a comparison with
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
accuracy in the perceptual identification task. The analysis on the accuracy and
latency data were submitted to two analyses of variance. The first was carried
out only on the trials that yielded a correct recognition of the target (indexed by
the verbal yes/no decision that participants had to perform after the button-press
response); the second was performed on all correct responses of the lexical
decision task, that is, regardless of whether the target word was later recognised
or not. The two analyses yielded similar results, although overall accuracy was
higher on the latter, and therefore only the analysis on the complete set of data
will be reported.
In the overall accuracy analysis, we found significant main effects of mask
type, F1(3, 240) = 9.8, MSE = 644.3, p < .001; F2(3, 156) = 10.6, MSE = 446.4, p
< .001, of duration, F1(3, 80) = 7.4, MSE = 2340.8, p < .001; F2(3, 156) = 63.2,
MSE = 527.9, p < .001, and of length, F1(1, 80) = 5.5, MSE = 694.7, p < .05;
F2(1, 52) = 6.4, MSE = 1141.4, p < .05. Separate analyses were run for the
different durations.
Table 3 show the results of the experiment for the brief duration condi-
tions. In the 16:33 ms duration, the analysis on the accuracy data showed a
reliable effect of mask type, F1(3, 63) = 3.84, MSE = 611, p < .01; F2(3, 156)
= 3.13, MSE = 437, p < .05. Planned comparisons revealed that higher accu-
racy was yielded in the vowel-preserving mask condition, t1(21) = 2.46, p <
.05, SE = 3.83, t2(55) = 3.00, p < .01, SE = 3.23, and in the consonant-
preserving mask condition, t1(21) = 2.47, p < .05, SE = 2.76, non significant
by items, not in the pseudohomophone condition, compared to the control
condition.
TABLE 3
Accuracy percentage in the lexical decision task (Experiment 3)
The mean latencies were 805, 754, 820, and 805 ms in the 16:33 condition
interval and 780, 797, 813, and 814 ms in the 33:33 condition interval, for the
same consonant, same vowel, pseudohomophone, and control conditions,
respectively. In the analysis of latencies of the short duration the mask effect
was significant by participants, F(3, 63) = 4.06, MSE = 12572, p < .01, showing
an advantage for the vowel-preserving, as compared to the consonant-preserving
mask condition, t1(21) = 2.09, SE = 37.9, p < .05. For the 33:33 ms duration no
effect was reliable.
Table 3 show the results of the analyses in the long duration conditions
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
(50:33 and 50:67 ms). In the analysis of correct responses five participants were
excluded in the 50:33 condition and two participants in the 50:67 condition. In
the 50:33 ms duration we found significant effects of mask type, F1(3, 42) =
2.97, MSE = 607.17, p < .05; F2(3, 156) = 3.78, MSE = 506.94, p < .05. Planned
comparisons revealed that accuracy was significantly lower in the control mask
condition as compared to the consonant-preserving mask condition, t1(14) = 4.1,
SE = 3.1, p < .01, t2(55) = 2.81, SE = 4.5, p < .01, the vowel-preserving mask,
t1(14) = 3.1, SE = 4.3, p < .01, t2(55) = 3.07, SE = 4.6, p < .01, and the
pseudohomophone mask, t1(14) = 2.45, SE = 5.0, p < .05, t2(55) = 2.15, SE =
4.8, p < .05. The comparison between the consonant-preserving mask and the
vowel-preserving mask was not significant. Mean latencies for the four condi-
tions were (in the same order as above): 733, 750, 739, 767 ms. The mask effect
was not significant.
In the 50:67 ms duration, the analysis of the accuracy data revealed a sig-
nificant effect of mask type, F1(3, 51) = 5.81, MSE = 748.39, p < .01; F2(3, 156)
= 7.9, MSE = 400.14, p < .001. Planned comparisons showed that accuracy for
the control mask condition was significantly lower than for the consonant-
preserving mask, t1(17) = 3.97, SE = 4.6, p < .01; t2(55) = 5.04, SE = 3.72, p <
.001, for the vowel-preserving mask, t1(17) = 3.11, SE = 5.3, p < .01; t2(55) =
3.78, SE = 4.66, p < .001, and for the pseudohomophone mask, t1(17) = 3.37, SE
= 5.4, p < .01; t2(55) = 3.76, SE = 4.04, p < .001. The comparison between the
consonant-preserving mask and the vowel-preserving mask was not significant.
Mean latencies were 727, 728, 713, and 764 ms for the different masking con-
ditions (in the same order as above). The analysis showed a significant masking
effect by participants, F1(3, 57) = 2.85, MSE = 3342.2, p < .05, F2(3, 165) = 1.03,
MSE = 182,720.3, p > .1, with an advantage for the pseudohomophone compared
to the control condition, t1(19) = 72.62, SE = 19.5, p < .05; t2(55) = 71.92, SE
= 33.2, p = .06.
Discussion
The results of the lexical decision task show an overall significant masking
reduction effect. At the shortest duration, both the same vowel and the same
consonant masks, but not the pseudohomophone mask, yielded higher accuracy
CONSONANTS AND VOWELS IN ITALIAN 421
compared to the control mask. We should also note that a very similar trend was
apparent in both accuracy and latency data.
In contrast to Experiments 1 and 2, no difference was found between the
same vowel and the same consonant masks. The absence of a difference between
the two mask types when the task is lexical decision might be interpreted dif-
ferently. It might indicate that the vowel advantage found at the short SOAs in
Experiments 1 and 2 is not reliable. Alternatively, it might indicate different
processing operations. Lexical decision might be conducted on the basis of an
orthographic code, and therefore if the vowel advantage is phonologically based,
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
it would not appear. If the first alternative is correct (due perhaps to some
confounding in the linguistic stimuli of Experiments 1±2), changing the mate-
rials would make it less likely to replicate the vowel advantage in the perceptual
identification task. Experiment 4 was conducted with this aim, and was carried
out using the same paradigm as in Experiment 1, i.e., the perceptual identifi-
cation task.
EXPERIMENT 4
In order to verify the effect of the serial position of the changed phoneme, the
position of the changed phoneme in words with different CV structures was
manipulated, in a way similar to that of Perry and Ziegler (2002). Thus, changes
of the consonant of the CCVCV structures were in the second position, whereas
they were in the third position in the CVCCV and CVCV masks. The vowel
change was in the second position in CVCCV and CVCV masks, and in third
position in CCVCV masks. This allowed both a contrast of consonant vs vowel
change, and a contrast of serial position (although in a limited range, second vs.
third position).
Method
Participants. Eighty-one students of the University of Padua participated in
the experiment. None had participated in the previous experiments.
words had been selected in a way that made possible their transformation in a
pseudohomophone, i.e., they included one of the Italian phonemes /i/, /k/, /u/
that could be spelled with the graphemes: y, k, w, respectively.
Four lists of 84 words each were constructed. In each list there were 28 target
words for each structure type, for a total of 84 words. Each target in the list was
associated to one of the four mask types, and for each mask type there were
seven different target words. In addition, six blank trials were also included,
formed by one of the nonword mask types (made from nonexperimental words)
presented without a target word. Each participant was presented with one of the
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
four lists of 84 trials, plus the 6 blank trials, for a total of 100 trials, at one of the
two exposure durations.
Results
Mean accuracy for each condition and each participant was calculated. Thirty-
two participants were discarded in the 16:33 SOA, and one in the 33:33 SOA
because their mean accuracy was below 10% of the total trials presented. Thus,
the analyses were carried out on the data for 20 participants for the 16:33 SOA,
and 28 for the 33:33 SOA.
The design included three levels of word structure, four types of mask, and
two SOAs (see Table 4). In the overall analysis there were significant effects of
TABLE 4
Accuracy percentage in the different conditions of Experiment 4
(perceptual identification)
Type of mask
SOA 16:33 ms
CVCCV 24.4 23.6 25.7 8.6
CCVCV 26.6 25.0 34.3 15.1
CVCV 14.3 20.0 22.2 10.7
Mean 21.7 22.8 27.4 11.5
SOA 33:33 ms
CVCCV 51.5 49.6 57.1 29.6
CCVCV 51.0 52.6 54.0 30.1
CVCV 47.5 47.5 51.0 31.8
Mean 50.0 49.9 54.1 30.5
CONSONANTS AND VOWELS IN ITALIAN 423
both mask type, F1(3, 138) = 36.04, MSE = 0.002, p < .001; F2(3, 243) = 33.82,
MSE = 0.03, p < .001, and of word structure, by participants only, F1(2, 92) =
4.93, MSE = 0.02, p < .01. The SOA was also significant, F1(1, 46) = 29.33,
MSE = 0.30, p < .001; F2(1, 81) = 262.33, MSE = 0.04, p < .001. Only the mask
by SOA interaction was marginally significant, F1(3, 138) = 2.34, MSE = 0.03,
p = .08; F2(3, 243) = 2.47, MSE = 0.03, p = .062.
In the 16:33 SOA condition there was a significant effect of mask type,
F1(3, 57) = 10.78, MSE = 0.008, p < .001; F2(3, 249) = 12.36, MSE = 0.03, p <
.001. Pairwise comparisons carried out on the different mask types showed that
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
the difference between each experimental condition and the control was sig-
nificant: t1(19) = 3.28, SE = .03, p < .01; t2(83) = 4.09, SE = .02, p < .001 for the
same consonant, t1(19) = 3.51, SE = .03, p < .01; t2(83) = 4.45, SE = .02, p = .001
for the same vowel, t1(19) = 5.44, SE = .03, p < .001; t2(83) = 5.99, SE = .03, p <
.001 for the pseudohomophone. The difference between same consonant and
pseudohomophone was also marginally significant, t1(19) = 71.93, SE = .03, p
< .06; t2(83) = 71.94, SE = .03, p = .06.
In the 33:33 SOA condition mask type was significant, F1(3, 81) = 34.46,
MSE = 0.009 , p < .001; F2(3, 249) = 27.7, MSE = 0.03, p < .001. The difference
between each experimental condition and the control was significant: t1(27) =
7.91, SE = .02, p < .001; t2(83) = 7.14, SE = .03, p < .001 for the same
consonant, t1(27) = 6.05, SE = .02. p < .001; t2(83) = 6.9, SE = .03, p = .001 for
the same vowel, t1(27) = 10.6, SE = .02, p < .001; t2(83) = 8.3, SE = .02, p < .001
for the pseudohomophone. No other difference was significant.
An analysis was carried out to investigate the extent of guessing attempted by
participants on the basis of the mask. No mask was accurately reported in the
short SOA, while in the 33:33 SOA mask report was 0.21%. For blank trials,
where the target was not presented, accurate report of the mask was 5.8% and
3.0% in the 16:33 and 33:33 exposures, respectively.
Further analyses were carried out in order to test for possible serial effects of
the position of the changed phoneme. These analyses were conducted on the
same consonant and same vowel masks only, with duration (2 SOAs) and
position of the changed phoneme (second vs. third letter/phoneme) as inde-
pendent variables. Thus, the data for the CVCCV and CVCV structures in the
same consonant condition and of the CCVCV in the same vowel condition
formed the condition ``phoneme change in second position'', while the data for
CVCCV and CVCV structures in the same vowel condition, and of the CCVCV
structures in the same consonant condition formed the ``phoneme change in
third position'' condition. The analysis with word structure, serial position, and
SOA as factors showed that only the effect of SOA, F1(1, 46) = 30.90, MSE =
0.173, p < .001; F2(1, 81) = 224.33, MSE = 0.03, p < .001, and word structure
were significant, by participants only, F(2, 92) = 3.53, MSE = 0.03, p < .05. No
other effect was significant.
424 COLOMBO ET AL.
Discussion
The data of Experiment 4 show a significant masking reduction effect for all the
mask conditions, sharing most of the letters/phonemes with the target word, as
compared to the control. However, the trend that was apparent in Experiments 1
and 2, of an advantage of the same vowel mask condition, was not replicated.
This result is consistent with Perry and Ziegler's (2002) failure to find evidence
for a consonant advantage in English. However, in the present data we also
failed to replicate the serial position effect that Perry and Ziegler found. As the
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
manipulation of serial position in the present study was limited, more stringent
tests may be necessary in order to draw strong conclusions on this issue.
GENERAL DISCUSSION
The present experiments were aimed at investigating whether the consonant
advantage found by Berent and Perfetti (1995) with a backward masking
paradigm and by Lee et al. (2001) with the missing letter paradigm for the
English language can be replicated in a language with a regular spelling±sound
mapping. In Experiments 1 and 2 we did not find any evidence for a consonant
advantage. Rather, the data showed an earlier computation of vowels. This result
however was not replicated in lexical decision. We argued that a failure to find a
vowel advantage in lexical decision might be due to the different characteristics
of the task (i.e., the response is based on a partial orthographic representation).
Thus, a replication of Experiment 1, with the same paradigm, and different
materials, might provide an answer to this ambiguous result. The overall results
show a significant masking reduction effect for conditions in which target and
mask share most of the letters/phonemes, but no reliable priority of either
consonant or vowel. This result is in contrast with that found for English by
Berent and Perfetti, and by Lee et al. (although, not by Perry & Ziegler, 2002,
and by Lukatela & Turvey, 2000).
A possible account of the data of the English and Italian experiments is that
the consonant advantage arises from the characteristics of the English language,
namely, the much higher degree of consistency of the spelling±sound mapping
for consonants compared to that for vowels. An independent motivation for an
explanation based on the different levels of grapheme±phoneme consistency for
consonants and vowels in English might come from recent connectionist models
of single word reading (e.g., Plaut, McClelland, Seidenberg, & Patterson, 1996;
Zorzi, Houghton, & Butterworth, 1998; see also the discussion in Colombo,
2000). Phonological assembly in these models is sensitive to the consistency of
the mapping from graphemes to phonemes. Inconsistent graphemes produce
competition among different phoneme candidates; this competitive (or ``clean
up'') process takes time and results in a slower activation of the inconsistent
phonemes. This is exactly the case for English vowels, but it does not apply to
Italian (where consonant and vowels are almost equally consistent).
CONSONANTS AND VOWELS IN ITALIAN 425
However, given the inconsistent results in the literature, one might wonder
whether the consonant advantage exists at all, even in English. On the one hand,
the processing advantage of consonants has been demonstrated by the studies of
two independent groups of researchers, Berent and Perfetti (1995) and Lee et al.
(2001), with different stimuli and different paradigms. Although the selection of
the materials in Berent and Perfetti's study has been criticised for a possible
confounding with serial position of the changed phoneme (Coltheart, 2000), this
criticism does not apply to the study by Lee et al. (2001). On the other hand, the
failure to replicate the consonant advantage by Lukatela and Turvey (2000) and
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
by Perry and Ziegler (2002) raise some doubts about the reliability of the effect.
Perry and Ziegler (2002) suggested that backward masking is not a suitable
paradigm for testing word recognition, because it is sensitive to guessing
strategies. In their experiments they indeed found evidence of guessing attempts
by participants. A possible interpretation is therefore that the consonant
advantage may reflect a greater facility to guess the identity of words on the
basis of consonants, rather than of vowels. This is actually a plausible inter-
pretation, given that consonants are more informative even under normal pre-
sentation conditions (Adams, 1981). However, the paradigm used by Lee et al.
(2001) is not subject to this criticism, because they measured gaze durations.
Therefore, the possibility remains that the consonant advantage is not, or not
only, a consequence of guessing strategies based on the greater informative
value of consonants, but is also dependent on reading mechanisms related to the
early computation of phonology.
In fact, the attribution by Perry and Ziegler (2002) of Berent and Perfetti's
(1995) results to the guessing strategies of participants in their experiments lends
itself to criticism. In the paradigm they used, the target word was preceded by a
pattern mask (a string of Xs). Thus, as both the target and the mask were forward
and backward masked, the target was less visible than the mask when its pre-
sentation duration was the same as the mask. That this is so is indicated by the
reported average accuracy in the 43:43 exposure condition of Experiment 1
(around 37% for the experimental conditions, 8% for the control condition),
which is lower as compared to the short mask exposure (43 for the target and 28
for the mask) in which the target duration is longer than the mask, and where
accuracy is around 53% for the experimental conditions, and 27% for the control
condition. And in fact the false mask report is higher in the long mask exposure:
around 14% in the experimental conditions of Experiment 1 and 26% in
Experiment 2, where also nonword targets were included, as compared to 2% of
the 43:28 exposure.
Perry and Ziegler (2002) consider the 43:43 exposure (together with the
inclusion of nonwords as targets) as critical for their interpretation of whether
the consonant advantage and/or serial effects depend on guessing strategies,
rather than being the consequence of ``normal'' reading mechanisms. They
assume that false mask reports are an indication of guessing strategies induced
426 COLOMBO ET AL.
Experiment 2. In the present experiments, on the other hand, the mask report
was practically absent, suggesting that guessing, at least on the basis of the
mask, was not influential. Whether guessing occurs at all, based on target
information, may be another matter deserving attention. In addition, several
other factors may contribute to the identification of words under these condi-
tions, including lexical factors (as suggested by Lee et al., 2001, in press).
In the experiments of the present study there was a pseudohomophone
masking reduction effect in both the spelling and the naming tasks, not in the
lexical decision task. Moreover, the pseudohomophone condition in general did
not show any significant advantage compared to the other experimental condi-
tions. A possible interpretation of this pattern of results is that the advantage for
the pseudohomophone at the phonological level (being pronounced like a word)
was nullified by the disadvantage at the orthographic level, so that the final
effect is the same as for masks sharing only consonants or vowels, when the
pseudoword had to be pronounced. This is because the pseudohomophone was
not graphemically equated to the control condition, either in graphemic simi-
larity, or in spelling±sound consistency. This fact should have made the masking
reduction effect (i.e., greater accuracy as compared to the control condition) less
easy to obtain. But in all the experiments (except lexical decision) at the short
duration there was a significant masking reduction effect. Thus, as the effect was
clear and robust in both perceptual identification and naming, the likely inter-
pretation is that it reflected phonological processing in those tasks. In contrast, if
lexical decision could be carried out on the basis of the output of early ortho-
graphic processing, there would be no advantage of accessing the lexical
phonology of a real word. Thus, the lack of a significant masking reduction
effect in lexical decision might reflect the early stages of graphemic activation.
In conclusion, the presence of clear and significant masking reduction effects
in both perceptual identification and naming is consistent with the idea that
participants carried out the tasks by relying on a phonological representation of
the masked words. Despite the evidence from neuropsychological and
developmental data in Italian for a distinct representation of consonants and
vowels, the present study found no evidence for an early emergence of the
phonological representations of consonants in a language with regular spelling±
sound correspondences, suggesting that the consonant advantage found by
CONSONANTS AND VOWELS IN ITALIAN 427
Berent and Perfetti and by Lee et al. was limited to a language with inconsistent
grapheme-to-phoneme mappings.
REFERENCES
Adams, M. J. (1981). What good is orthographic redundancy? In O. J. L. Singer & H. Tzeng (Eds.),
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
De Mauro, T., Mancini, F., Vedovelli, M., & Voghera, M. (1993). Lessico di frequenza dell'italiano
parlato. Milan: Etaslibri.
Ferrand, L., & Grainger, J. (1992). Phonology and orthography in visual word recognition: Evidence
from masked nonword priming. Quarterly Journal of Experimental Psychology, 45A, 353±372.
Ferrand, L., & Grainger, J. (1993). The time course of orthographic and phonological code activation
in the early phases of visual word recognition. Bulletin of the Psychonomic Society, 31, 119±122.
Ferrand, L., & Grainger, J. (1994). Effects of orthography are independent of phonology in masked
form priming. Quarterly Journal of Experimental Psychology, 47A, 365±382.
Forster, K. I., & Davis, C. (1991). The density constraint on form-priming in the naming task:
Interference effects from a masked prime. Journal of Memory and Language, 30, 1±25.
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
Grainger, J., & Ferrand, L. (1994). Phonology and orthography in visual word recognition: Effects of
masked homophone primes. Journal of Memory and Language, 33, 218±233.
Lee, H. W., Rayner, K., & Pollatsek, A. (2001). The relative contribution of consonants and vowels
to word identification during reading. Journal of Memory and Language, 44, 189±205.
Lee, H. W., Rayner, K., & Pollatsek, A. (in press). The processing of consonants and vowels in
reading: Evidence from the fast priming paradigm. Psychonomic Bulletin and Review.
Lukatela, G., & Turvey, M. T. (1994a). Visual lexical access is initially phonological: 1. Evidence
from associative priming by words, homophones, and pseudohomophones. Journal of Experi-
mental Psychology: General, 123, 107±128.
Lukatela, G., & Turvey, M. T. (1994b). Visual lexical access is initially phonological: 2. Evidence
from phonological priming by homophones and pseudohomophones. Journal of Experimental
Psychology: General, 123, 331±353.
Lukatela, G., & Turvey, M. T. (2000). An evaluation of the two-cycles model of phonology
assembly. Journal of Memory and Language, 42, 183±207.
Lupker, S. J. , Brown, P., & Colombo, L. (1997). Strategic control in a naming task: Changing routes
or changing deadlines? Journal of Experimental Psychology: Learning, Memory and Cognition,
23(3), 570±590.
McCann, R. S., & Besner, D. (1987). Reading pseudohomophones: Implications for models of
pronunciation and the locus of the word-frequency effect in word naming. Journal of Experi-
mental Psychology: Human Perception and Performance, 13, 14±24.
McCann, R. S., Besner, D., & Davelaar, E. (1988). Word recognition and identification: Do word-
frequency effects reflect lexical access? Journal of Experimental Psychology: Human Perception
and Performance, 14, 693±706.
Monsell, S., Patterson, K. E., Graham, A., Hughes, C. H., & Milroy, R. (1992). Lexical and sub-
lexical translation of spelling to sound: Strategic anticipation of lexical status. Journal of
Experimental Psychology: Learning, Memory and Cognition, 18, 452±467.
Naish, P. (1980). The effects of graphemic and phonemic similarity between targets and masks in a
backward visual masking paradigm. Quarterly Journal of Psychology, 32, 57±68.
Paap, K. R., & Noel, R. W. (1991). Dual route models of print to sound: Still a good horse race.
Psychological Research, 53, 13±24.
Perfetti, C. A., & Bell, L. C. (1991). Phonemic activation during the first 40 ms of word identifi-
cation: Evidence from backward masking and masked priming. Journal of Memory and Lan-
guage, 30, 473±485.
Perfetti, C. A., Bell, L. C., & Delaney, S. M. (1988). Automatic (prelexical) phonemic activation in
silent word reading: Evidence from backward masking. Journal of Memory and Language, 27,
59±70.
Perry, C., & Ziegler, J. C. (2002). On the nature of phonological assembly: Evidence from backward
masking. Language and Cognitive Processes, 17(1), 31±59.
Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. E. (1996). Understanding normal
and impaired word reading: Computational principles in quasi-regular domain. Psychological
Review, 103, 56±115.
CONSONANTS AND VOWELS IN ITALIAN 429
Rastle, K., & Coltheart, M. (1999). Serial and strategic effects in reading aloud. Journal of
Experimental Psychology: Human Perception and Performance, 25, 482±503.
Rayner, K., Pollatsek, A., & Binder, K. (1998). Phonological codes and eye movements in reading.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 476±497.
Stone, G. O., Vanhoy, M., & Van Orden, G. C. (1997). Perception is a two-way street: Feedforward
and feedback phonology in visual word recognition. Journal of Memory and Language, 36, 337±
359.
Treiman, R., Mullenix, J., Bijeljac-Babic, R., & Richmond-Welty, E. D. (1995). The special role of
rimes in the description, use, and acquisition of English orthography. Journal of Experimental
Psychology: General, 124, 107±136.
Downloaded by [University of Tennessee, Knoxville] at 19:28 02 January 2015
Van Orden, G. C. (1987). A ROWS is a ROSE: Spelling, sound, and reading. Memory and Cognition,
15, 181±198.
Van Orden, G. C., Johnston, J. C., & Hale, B. L. (1988). Word indentification in reading proceeds
from spelling to sound to meaning. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 14, 371±386.
Van Orden, G. C., Pennington, B. F., & Stone, G. O. (1990). Word identification in reading and the
promise of subsymbolic psycholinguistics. Psychological Review, 97, 488±522.
Ziegler, J. C., & Jacobs, A. M. (1995). Phonological information provides early sources of constraint
in the processing of letter strings. Journal of Memory and Language, 34, 567±593.
Ziegler, J., Van Orden, G. C., & Jacobs, A. M. (1997). Phonology can help or hurt the perception of
print. Journal of Experimental Psychology: Human Perception and Performance, 23, 845±860.
Zorzi, M. (2000). Serial processing in reading aloud: No challenge for a parallel model. Journal of
Experimental Psychology: Human Perception and Performance, 26, 847±856.
Zorzi, M., Houghton, G., Butterworth, B. (1998). Two routes or one in reading aloud? A connec-
tionist dual-process model. Journal of Experimental Psychology: Human Perception and Per-
formance, 24, 1131±1161.
APPENDIX
List of target words and pseudoword masks used in
Experiments 1±3*
Target Same consonant Same vowel Pseudohomophone Control
(Continued)
430 COLOMBO ET AL.
(Continued)
432 COLOMBO ET AL.
(Continued)
CONSONANTS AND VOWELS IN ITALIAN 433
CVCV words and position of changed phoneme (Continued)