The Spanish Journal of Psychology
Copyright 2008 by The Spanish Journal of Psychology
2008, Vol. 11, No. 1, 26-35
Children Like Dense Neighborhoods: Orthographic
Neighborhood Density Effects in Novel Readers
Jon Andoni Duñabeitia1 and Eduardo Vidal-Abarca2
1Universidad de La Laguna
2Universitat de València
Previous evidence with English beginning readers suggests that some orthographic effects, such as the
orthographic neighborhood density effects, could be stronger for children than for adults. Particularly,
children respond more accurately to words with many orthographic neighbors than to words with few
neighbors. The magnitude of the effects for children is much higher than for adults, and some researchers
have proposed that these effects could be progressively modulated according to reading expertise. The
present paper explores in depth how children from 1st to 6th grade perform a lexical decision with words
that are from dense or sparse orthographic neighborhoods, attending not only to accuracy measures, but
also to response latencies, through a computer-controlled task. Our results reveal that children (like adults)
show clear neighborhood density effects, and that these effects do not seem to depend on reading expertise.
Contrarily to previous claims, the present work shows that orthographic neighborhood effects are not
progressively modulated by reading skill. Further, these data strongly support the idea of a general
language-independent preference for using the lexical route instead of grapheme-to-phoneme conversions,
even in beginning readers. The implications of these results for developmental models in reading and for
models in visual word recognition and orthographic encoding are discussed.
Keywords: lexical access; reading development; orthographic neighborhood; density effect
La investigación previa con lectores principiantes de ingles sugiere que algunos efectos ortográficos,
tales como los efectos de la densidad (vecindad ortográfica), podrían ser más fuertes para los niños que
para los adultos. En especial, los niños responden con mayor precisión a las palabras con muchos
vecinos ortográficos que a las palabras con pocos vecinos. La magnitud de los efectos para los niños
es mucho más alta que para los adultos, y algunos investigadores han propuesto que estos efectos
podrían modularse progresivamente en función de la competencia lectora. Este estudio explora en
profundidad cómo los niños de 1º a 6º curso llevan a cabo una decisión léxica con las palabras procedentes
de vecindades ortográficas densas o escasas, atendiendo no sólo a las medidas de precisión sino también
a las latencias de respuesta, mediante una tarea controlada por ordenador. Nuestros resultados revelan
que los niños (como los adultos) muestran claros efectos de densidad (vecindad ortográfica), y que
dichos efectos no parecen depender de la competencia lectora. Al contrario de observaciones previas,
el trabajo actual muestra que los efectos de vecindad ortográfica no se modulan progresivamente según
la competencia lectora. Además, estos datos claramente apoyan la idea de la preferencia por la ruta
léxica, que no depende del lenguaje, en vez de las conversiones grafema-a-fonema, incluso en lectores
principiantes. Se comentan las implicaciones de estos resultados para los modelos evolutivos de la
lectura y para los modelos de reconocimiento visual de las palabras y la codificación ortográfica.
Palabras clave: acceso léxico, evolución lectora, vecindad ortográfica, efecto de densidad
The research reported in this article has been partially supported by Grants SEJ2004-07680-C02-02/PSIC and SEJ2006-09238/PSIC
from the Spanish Government, and by Grant BFI05.310 from the Basque Government.
Address correspondence concerning this article to Jon Andoni Duñabeitia, Departamento de Psicología Cognitiva, Universidad de
La Laguna, 38205 - Tenerife (Spain). Phone: +34678635223. Fax: +34922317461. E-mail: firstname.lastname@example.org
ORTHOGRAPHIC EFFECTS IN CHILDREN
How does a reader access the meaning of a visually
neighbors (e.g., trial-trail). Substitution neighbors are the
presented single word? For decades, researchers in
ones that have been studied in more depth, and we will
psycholinguistics have been attempting to shed light on this
focus on them in the present research.
issue. By now, what seems clear is that there are various
processes that occur (or co-occur) when a reader faces a
printed word: Letter position and identity encoding (e.g.,
Substitution Neighborhood Density Effects
Grainger, Granier, Farioli, Van Assche, & van Heuven, 2006),
affix stripping and morphological decomposition (e.g.,
The index N is typically used to refer to the number of
Duñabeitia, Perea, & Carreiras, 2007; Frost, Kugler, Deustch,
substitution neighbors of a given word (see Coltheart et al.,
& Forster, 2005), lexeme processing (e.g., Pollatsek, Hyönä,
1977; Andrews, 1989, 1992). Words vary widely in the
& Bertram, 2000) and semantic integration (e.g., Shelton &
number of substitution neighbors that can be created by
Martin, 1992), among others. However, it still unknown to
changing a single letter. On the one hand, there are words
what extent orthographic processes interact with
with dense neighborhoods such as the word SAND, which
morphological or semantic processes. Several models and
has many substitution neighbors. On the other hand, there
theories support a localist approach, whereas others support
are words with sparse neighborhoods, such as BOTTLE,
fully distributed frameworks (for a recent review see Page,
which only has the word battle as a substitution neighbor.
2000). It is now known that even though a single word is
Laxon, Coltheart, and Keating (1988) designed them as
presented, and the reader may effectively achieve its
‘friendly’ words (the firsts) and ‘unfriendly’ words (the
meaning, many other forms and meanings are activated
latter). Finally, one can also find words with no substitution
while processing the target word. The present paper is
neighbors, called ‘hermits’ (e.g., LYNX).
centered on one specific orthographic relationship between
Generally, it has been assumed that the higher the density
word forms, which has been a focus of debate in recent
of N is for a word, the faster a response is given on that
decades: the orthographic neighborhood.
word, and the greater the accuracy (Andrews, 1989; Perea,
Coltheart, Davelaar, Jonasson, and Besner (1977)
1998; Pollatsek, Perea, & Binder, 1999; see Siakaluk, Sears,
characterized the orthographic neighborhood of a given word
& Lupker, 2002, for a review). This facilitation is generally
(referred to as N) as all the existing words that could be
obtained only in lexical decision tasks, and changing the
created by replacing one of its letters by another different
task sometimes implies a reversal of the effect, becoming
one. For example, the word SAND is said to have 11
inhibitory instead of facilitative. For example, words with
orthographic neighbors, since there are 11 words that share
dense neighborhoods tend to be recognized slower in
with it 3 letters in the same position (e.g., land, hand, band,
progressive demasking tasks (Carreiras, Perea, & Grainger,
send, said, sang …). There is robust evidence from diverse
1997; Grainger & Segui, 1990), speeded lexical decision
experimental tasks and paradigms showing that the
tasks and perceptual identification tasks (Snodgrass &
presentation of SAND activates the orthographic, lexical and
Mintzer, 1993), in on-line sentence reading monitored by
even the semantic representations of its neighbors (see
eye-tracking systems (Pollatsek, Perea, & Binder, 1999),
Siakaluk, Sears, & Lupker, 2002, for a review; also
and even when electrophysiological measures are considered
Duñabeitia, Carreiras, & Perea, in press, or Forster & Hector,
(see Holcomb, Grainger, & O’Rourke, 2002, for a larger
2002, for semantic activation of neighbors). Further, recent
amplitude of the N400 for words with high N density).
evidence has revealed that not only are these orthographic
Carreiras and colleagues (1997) explained these cross-task
neighbors activated while accessing a word, but also other
differences in N density effects in the following way:
types of neighbors, such as addition or deletion neighbors,
“Effects on orthographic neighborhood in the various tasks
or the neighbors by transpositions. Davis and Taft (2005;
can vary from being facilitative to being inhibitory as a
also Davis & Perea, 2007) showed that words created by
function of the extent to which participants base their
adding or deleting a letter from a base word are also
responses on unique word identification” (p. 868).
activated during lexical retrieval (e.g., stand and sad are
Recently Lavidor, Johnston and Snowling (2006) compiled
addition/deletion neighbors from SAND). Moreover, there
the most robust previous data in this line, and gave theoretical
is evidence in favor of a co-activation, not only of words
explanations for the (sometimes) facilitative effect of
that share all letters but one in the same position, but also
substitution neighbor density in lexical decision tasks. They
of words that share all letters but in different positions, like
argued that both the Interactive Activation frameworks (e.g.,
the transposed-letter neighbors (e.g. trial and trail; see
McClelland & Rumelhart, 1981) and the Multiple Read Out
Andrews, 1996; Chambers, 1979). Therefore, the general
model (Grainger & Jacobs, 1996) could readily account for
term ‘orthographic neighbor’ introduced by Coltheart and
these data. Firstly, the Interactive Activation models assume
colleagues, has been harmonized with a more detailed
that feedback activation from word to letter level facilitates
classification, depending on the overlap between the words:
processing, and therefore, the more activated candidates in
substitution neighbors (e.g., sand-land), addition/deletion
the lexicon (namely neighbors), the more top-down feedback
neighbors (e.g., sad-sand-stand) or transposition-letter
activation could be expected (see Andrews, 1989, 1997, for
DUÑABEITIA AND VIDAL-ABARCA
a similar approach). Secondly, the Multiple Read Out model
(2002) studied children from 5 to 7 years old in a naming
accounts for this facilitative effect in terms of the increased
task, and partially replicated the previous results, showing
lexical activation that helps the decision stage in the lexical
that children were less accurate with words from sparse
decision task: “words with a large N lead to the partial
neighborhoods (see their Experiment 2 notwithstanding).
activation of a large number of word representations in
The fact that children do show robust N density effects
memory” (Holcomb, Grainger, & O’Rourke, 2002, p. 939).
was explained in terms of an early preference of beginning
This last view is totally in line with explanations of how
readers to use a lexical route, avoiding grapheme-to-phoneme
lexical decisions are carried out, based on the summed
conversion by using cues of orthographic segments (see Frith,
activation of all positively activated word representations (see
1980, 1985). In fact, other researchers have reached similar
Balota & Chumbley, 1984). Other models of word recognition,
conclusions when attending to different orthographic effects
however, have more difficulties in accounting for the N density
in children. Perea and Estévez (2008) showed that beginning
effect in lexical decision tasks. For example, the Search model
readers tend to commit more lexicalizations than adult readers
(Forster, 1976) and the Activation-Verification framework
in a naming task where nonwords created by letter
(Paap et al., 1982), predict that having many substitution
transpositions were involved (e.g., CHOLOCATE; see Perea
neighbors should increase reaction times (but see Forster &
& Fraga, 2006, or Perea & Lupker, 2003, for a review on
Shen, 1996, and Paap & Johansen, 1993).
transposed-letter effects). Second grade children lexicalized
Lavidor and colleagues (2006) also reviewed the
more nonwords than 4th grade students (e.g., pronounced
evidence regarding the hemispheric distribution of the N
CHOCOLATE when visually presented CHOLOCATE), and
effects. Recent data have shown that the N effect is more
the latter group lexicalized much more than a group formed
robust in the left visual field than in the right visual field
by college students (percentages of lexicalizations were 45%,
(note that the left visual field relies on the right hemisphere,
39% and 30%, respectively). Sebastián-Gallés and Parreño
whereas the right visual field relies on the left hemisphere;
(1995) also found a similar pattern of orthographic effects
see Rubinstein, Henik, & Dronkers, 2001; also Lavidor &
in children. These authors showed that at an early reading
Walsh, 2003). This lateralized effect perfectly matches
stage novel readers tend to [over] use a lexical route, as
Coltheart’s right hemisphere reading hypothesis (1980, 2000),
deduced from the lexicalization errors they committed when
and has been explained by the poorer resolution and focusing
reading nonwords like ABOGEDO, which they articulated
of the right hemisphere, that relies more on supra-letter units.
as ABOGADO (the Spanish word for lawyer).
The repercussion of this lateralization effect has been huge,
However, it has been recently proposed that the use of
and recent models on letter encoding have echoed it
the grapheme-to-phoneme conversion rules and the use of
(SERIOL model by Whitney, 2001; Split-Fovea model by
lexical direct routes may vary from language to language.
Shillcock, Ellison, & Monaghan, 2000).
Specifically, it has been said that beginning readers from
transparent languages where the grapheme-phoneme
correspondence is carried out almost in a one-to-one fashion
N Effects in Special Populations
might preferably use this mechanism rather than a lexical
route, whereas novel readers from more opaque orthography-
The theoretical implications of a well-defined N density
to-phonology conversions might develop preferences for
effect are noteworthy, as are the applied educational
using the lexical pathway (Ziegler & Goswami, 2006). These
implications of this effect. Curiously, however, N density
authors propose that “children who are learning to read more
effects have been typically studied in adult normal reader
orthographically consistent languages, such as Greek,
populations, and only a few studies have concentrated their
German, Spanish or Italian, rely heavily on grapheme-
efforts on uncovering how children or older adults reflect
phoneme recording strategies because grapheme-phoneme
these effects. To our knowledge, only one study focused on
correspondences are relatively consistent” (p.431). There is
neighborhood density effects and their relation with aging,
a clear discrepancy between this assumption and the previous
demonstrating that older adults do not show them (Spieler
data in favor of lexical route predominance in young readers
& Balota, 2000). Rather more copious is the evidence about
from transparent/consistent languages (Perea & Estévez,
children’s orthographic processing and N effects. For
2008; Sebastián-Gallés & Parreño, 1995). The present work
instance, Laxon, Coltheart and Keating (1988) presented a
will try to shed some light on this debate.
list of words and nonwords of dense and sparse
This study was conducted in order to explore the scope
neighborhoods to children in a naming and in a lexical
of the substitution neighborhood density effects in children.
decision task. Correct responses on the items were measured
There is previous evidence revealing that beginning readers
for a group of children from 2nd and 3rd grade. Their results
do show great N effects (Laxon et al., 1988; Laxon,
showed that ‘friendly’ words were read more accurately than
Masterson, & Moran, 1994), but there is still a gap to be
‘unfriendly’ words, both in naming and lexical decision
filled regarding the gradation of these effects, if there is
(with more than a 10% accuracy difference in both cases).
some. While Laxon and colleagues only tested 2nd and 3rd
In a subsequent study, Laxon, Gallagher and Masterson
grade students, the present paper reports evidence with
ORTHOGRAPHIC EFFECTS IN CHILDREN
groups of children from all the six grades of Primary School.
analyzed using the B-PAL software (Davis & Perea, 2005)
It is extremely important to uncover if beginning readers
that provides the most common psycholinguistic indexes.
and skilled readers do show different patterns in the
Twenty two of these words had a dense neighborhood,
magnitude of N effects, because, if so, models of reading
whereas the other 22 had a sparse neighborhood. Words in
development and models of orthographic encoding should
the dense neighborhood condition had a mean frequency of
echo these differences.
94.4 appearances per million words (standard deviation:
Regarding the experimental paradigms used in the
131.9), and a mean number of 4.8 letters (± 1.5). Their
previous studies, the present study has a clear advantage in
bigram token frequency was 770.1 (± 555.7), and their
comparison with them. While Laxon and collaborators
bigram type frequency was 37.5 (± 35.6). The mean number
restricted their analyses to the accuracy data, in the present
of neighbors of these words was 5.6 (range: 2-11). For
study we report analyses of both reaction time measures
example, the Spanish word MENTA (meaning mint) has 8
and accuracy measures. It is a fact that in many studies with
different orthographic neighbors that can be created just by
adult samples, the N density effects only show up when
changing one of the letters from the original string (i.e.,
measuring response latencies, and that accuracy data do not
lenta, renta, manta, venta, or mente, translated as slow, rent,
always reveal the same effects (Carreiras et al., 1997;
blanket, sale and mind, respectively). On the other hand,
Pollatsek et al., 1999; Sears, Hino, & Lupker, 1995).
words in the sparse neighborhood condition had a mean
Moreover, with the appropriate analyses, reaction times with
frequency of 88.1 (± 92.9) appearances per million words,
children can be very informative, as they are for adults
and a mean length of 5.8 letters (± 1). Their mean bigram
(Perea & Algarabel, 1999).
token frequency was 726.2 (± 369.7) and their mean bigram
In Laxon’s studies, children were presented with strings
type frequency was 48.5 (± 26.5). The mean number of
of letters printed on cards, and they had to either pronounce
orthographic neighbors of these words was 0.5 (range: 0-
them, or to make a lexical judgment about them. The
1). An example from this subset of words is RURAL, that
experiment that we report was conducted using an
only has one orthographic neighbor (mural). As can be seen,
automatized computer-controlled random presentation of the
the two sets were closely matched in all statistics except for
stimuli, instead of using more course presentation methods
orthographic neighborhood, as this is the critical manipulation
that could be influenced by and confounded with context-
in the current experiment. We conducted t-tests to compare
dependent factors. The present study focuses on real word
the statistics of the different frequency measures for words
processing and on how the orthographic overlap between
in the sparse and dense neighborhood conditions and all the
words interferes with lexical access, and therefore no specific
comparisons resulted non-significant (all ps > .12). However,
manipulations were carried out on the nonwords. We decided
the critical comparison between the two subsets of words,
to avoid nonword manipulations also because, as stated
namely the difference in the number of orthographic
earlier, there is previous evidence showing that the existence
neighbors, resulted significant (t = 8.88, p < .01). Moreover,
of the N density effect is sometimes constrained to specific
words in the dense neighborhood and in the sparse
manipulations on nonwords.
neighborhood conditions had similar syllabic structures.
These two subsets of words formed an item block, which
was included in a list with another 31 Spanish words with
similar frequencies and length, and a neutral number of
neighbors (approximately 3), that were used for a reading
test standardization purpose (Duñabeitia, Vidal-Abarca, &
Izquierdo, 2008). In order to make the lexical decision
Two hundred and sixty-two children from primary
possible, a set of 75 nonwords matched in length to the
schools in Valencia took part in the present data collection.
words was also included. These nonwords were all
All of them had either normal or corrected-to-normal vision,
pronounceable, and did not comprise any illegal bigram or
and were native Spanish speakers. None of the participants
trigram (e.g., JUENO). All the stimuli were inserted in a
had learning disabilities, or any remarkable reading difficulty.
single list, and were randomly presented to the participants,
The sample included children from all levels of primary
so that no order repetition effects could be expectable.
school: 28 students from 1st grade, 33 from 2nd, 45 from
3rd, 40 from 4th, 60 from 5th, and 56 from 6th. They were
all tested at the end of their current course.
Participants were tested individually, in a silent and
well-lit room. All the data collection was carried out during
school hours, with permission from parents and teachers.
The experimental set comprised a block of 44 Spanish
The list of items was presented on a laptop computer with
words taken from the LEXESP database (Sebastián-Gallés,
a LCD monitor. The responses were recorded using the
Martí, Carreiras, & Cuetos, 2000). These words were
DMDX software for stimuli presentation and data collection
DUÑABEITIA AND VIDAL-ABARCA
(Forster & Forster, 2003). All the stimuli were presented
in the centre of the screen, in uppercase white 12 pt. Courier
New font, with a black background. Each trial consisted in
Incorrect or null responses (YES responses to nonwords,
the appearance of a fixation point (‘+++’) for 1200 ms,
NO responses to words, or lacks of response) were not
followed by the presentation of the target word, displayed
included in the latency data analyses. All the response times
for 5000 ms, or until a response was made by the
above or below the 2 standard deviation cutoffs were
participant. Participants were instructed to decide if the
triggered out for the analyses. Figure 1 summarizes the
string presented in the screen was, or was not, a legal
reaction times and error rates associated to the experimental
Spanish word, by pressing one of the two buttons. They
set of each of the groups of children in a graphical report.
were also told to do so as fast and as accurately as possible.
Instead of using the arithmetical mean, we chose the median
Participants’ response latencies were measured from the
for the analyses, since there is empirical evidence showing
target string’s onset to the button pressing. Responses were
that this is the best option for contaminated distributions
made by pressing ‘Z’ or ‘M’ buttons, labeled as ‘YES’ and
(see Ratcliff, 1993; Perea & Algarabel, 1999; see also Acha
‘NO’ respectively. Although no more buttons were labeled,
& Perea, in press, for a study with beginning readers).
those buttons surrounding the critical ‘Z’ and ‘M’ in the
ANOVAs were conducted for participant (F1) and item (F2)
qwerty keyboard were also programmed for collecting
median response times, based on a 6 (Grade: 1st, 2nd, 3rd,
responses (e.g., the ‘A’, ‘S’ and ‘X’ buttons for ‘NO’
4th, 5th, and 6th) x 2 (Neighborhood Density: Dense, Sparse)
responses, and the ‘N’, ‘J’ and ‘K’ buttons for ‘YES’
design. MinF values are also provided (Clark, 1973).
responses). This was done because our previous experience
with very young children strongly ensures that many
incorrect responses in lexical decision experiments are due
to errors in button pressing. At the beginning of the
ANOVAs on the response latencies of the words showed
experiment, a trained experimenter carefully read and
a main effect of Grade, revealing that children from higher
explained the instructions to the participants. All the
educational levels responded faster to the stimuli than children
participants carried out a practice with a list of 6 stimuli
from lower levels, F1(5, 256) = 53.48, p < .01, 1-? = 1;
(3 words and 3 nonwords), in order to get them used to the
F2(5, 263) = 135.37, p < .01, 1-? = 1; minF(5, 433) = 38.34,
p < .01. Also, the main effect of Neighborhood Density was
Figure 1. Lexical decision times in medians (in ms) and percentage of errors for words. Grey bars refer to the reaction times in the dense
neighborhood condition. Black bars refer to the reaction times in the sparse neighborhood condition. The line with triangles refers to the
percentages of errors in the dense neighborhood condition. The dotted-line with squares refers to the percentages of error in the sparse
ORTHOGRAPHIC EFFECTS IN CHILDREN
significant, F1(1, 256) = 131.57, p < .01, 1-? = 1; F2(1,
p < .01, 1-? = 1; F2(5, 263) = 7.16, p < .01, 1-? = .99;
263) = 10.96, p < .01, 1-? = .91; minF(1, 306) = 10.11, p
minF(5, 341) = 4.48, p < .01. Importantly, words in the dense
< .01. This effect showed that words in the dense
neighborhood condition were responded to more accurately
neighborhood condition were responded to faster than words
than words in the sparse neighborhood condition (a 3.8%
in the sparse neighborhood condition (130 ms faster)1 . The
difference), F1(1, 256) = 75.92, p < .01, 1-? = 1; F2(1, 263)
interaction between the two factors was only significant in
= 4.68, p < .0.3, 1-? = .58; minF(1, 295) = 4.41, p < .04.
the analysis by participants, and not in the analysis by items,
A strong Neighborhood Density effect was observed in the
nor in the minF, F1(5, 256) = 28.65, p < .01, 1-? = 1; F2(5,
present experiment. The interaction between this effect and the
263) = 1.84, p = .11, 1-? = .62; minF(5, 297) = 1.73, p =
educational level of the participants was only significant in the
.13. Planned comparisons were carried out for each level of
analysis by participants, and therefore, one could not argue that
Grade: First grade students, t1(27) = 9.00, p < .01; t2(42) =
the effect is totally different for younger than for elder scholars.
2.25, p < .03; Second grade students, t1(32) = 2.85, p < .01;
If one looks at the pattern for the 90, 70, 50, 30, and 10%
t2(42) = 1.64, p = .11; Third grade students, t1(44) = 1.35,
quantiles of the distributions (see Figure 2), it seems clear that
p = .18; t2(42) = .87, p = .39; Fourth grade students, t1(39)
all the groups show a shift of the entire reaction time distribution
= 2.63, p < .02; t2(42) = 1.15, p = .26; Fifth grade students,
for the dense neighborhood and sparse neighborhood conditions,
t1(59) = 4.11, p < .01; t2(42) = 1.36, p = .18; Sixth grade
in a very similar fashion: the shifts for the 2nd-6th grade students
students, t1(55) = 2.20, p < .04; t2(42) = .13, p = .90.
are almost identical, showing a virtually equal effect. The
children from 1st grade are the only ones showing a
disproportionate N density effect; whereas the rest show similar
patterns and magnitudes. However, when we performed a
Regarding the percentages of errors associated to the
logarithmic transformation of these data, and therefore removed
words, the Grade factor resulted in a significant effect,
the proportional increase between groups, the interaction that
showing that students from higher levels committed less
resulted significant (in the analysis by participants) was no
errors than students from lower levels, F1(5, 256) = 33.87,
longer significant (p > .09)2.
Figure 2. Group response time distributions for the experimental pairs. The circles represent the 10%, 30%, 50% (in bold), 70%, and
90% quantiles. These values were computed by computing the quantiles for individual participants and then averaging the computed
values for each quantile over the participants.
1 Even though we conducted the analyses using the median values instead of the arithmetical mean, the effects reported in this section
did not differ widely when the same analyses were conducted with the arithmetical mean. The main effect of Neighborhood Density was
significant, F1(1, 256) = 179.03, p < .01; F2(1, 263) = 14.84, p < .01; minF(1,3 06) = 13.70, p < .01. Also the effect of Grade was significant
when analyzing the data with arithmetical mean, F1(5, 256) = 48.82, p < .01; F2(5, 263) = 149.61, p < .01; minF(5, 408) = 36.80, p < .01.
2 We thank an anonymous reviewer for suggesting this analysis.
DUÑABEITIA AND VIDAL-ABARCA
The present results claim against a progressive gradation
of the neighborhood density effects, because, except for 1st
The present findings are clear-cut. First, we have been
grade readers, the other children show approximately similar
able to replicate the substitution neighborhood density effect
patterns, not only in reaction times, but also in error rates.
with beginning readers. Second, we have shown that this
It seems clear that only children from 1st grade show around
effect is not progressively degraded, but that, except for
a 10% difference effect in accuracy (similarly to the results
children in the 1st grade, the other readers do show virtually
from Laxon and colleagues), whereas the rest of the groups
the same pattern of effects. Third, the present work reveals
show a relatively low but significant effect, which resembles
that children not only show N effects in their response
the effect in adult samples. The same comparison with
accuracy, but that these effects can be efficiently captured
reaction times yield to an identical conclusion: only 1st grade
by reaction times. And fourth, our data reveal that
students show a huge reaction time effect, whereas the rest
orthographic effects in children can be satisfactorily studied
of children show more similar patterns. Therefore, it could
by mimicking the tasks and paradigms that are commonly
be that the same underlying mechanisms responsible for the
used with adult readers.
N density effects in adults are also responsible for the effects
Adult readers generally respond faster and more accurately
in children, and that it is not reading fluency or lexicon size,
to words with many substitution neighbors (e.g., SAND) than
nor expertise itself, that is modulating these effects. One
to words with few or no substitution neighbors (e.g., LYNX).
plausible explanation for integrating Laxon’s results and our
These neighborhood density effects have been largely
data could be that the children they tested were not proficient
investigated and replicated (see Andrews, 1989, 1992, 1997;
enough, and that, as do our children from 1st grade, they
Carreiras et al, 1997; Grainger & Jacobs, 1996; Perea, 1998).
present great variability associated to the high number of
These effects have yielded numerous theoretical consequences,
errors committed (note that in Laxon et al., 1988, the
and many models of visual word recognition have had to
participants failed in more than 20% of the trials,
implement changes in their frameworks in order to correctly
irrespectively of the neighborhood density condition, as was
capture N effects (e.g., Interactive Activation model by
the case in the younger readers from our experiment). In
McClelland & Rumelhart, 1981; Parallel Distributed
addition, we conducted an Age of Acquisition norming study
Processing model by Rumelhart & McClelland, 1986;
with the materials in the present experiment. Twenty-eight
Multiple Read Out model by Grainger & Jacobs, 1996; Dual
undergraduate students from the University of La Laguna
Route Cascade model by Coltheart, Rastle, Perry, Ziegler, &
and the University of Valencia rated each word in a 1-to-7
Langdon, 2001). Similarly, recent models of orthographic
scale (each punctuation referring to the age range in which
encoding have been satisfactorily designed, including specific
they though a given word was learnt). Words in Dense
mentions of these N effects (e.g., SOLAR model by Davis,
Neighborhoods were rated with a mean of 2.9 points, whereas
1999; SERIOL model by Whitney, 2001; Overlap model by
words in Sparse Neighborhoods were rated with a mean of
Gómez, Ratcliff, & Perea, 2007). However, the evidence with
3.7 points (p < .01). Interestingly, the age values associated
other populations, such as children, is scarce, even though
to each of these punctuations were 6 and 7 years respectively.
the implications for models in reading acquisition are vast.
Therefore, a possible explanation for the disproportionate N
Previous evidence has shown that neophyte readers do
effect shown in the 1st graders could be reflecting the fact
show neighborhood density effects (Laxon et al., 1988;
that some of the words in the Sparse Neighborhoods were
Laxon et al., 2002; Laxon et al., 1994). These studies mainly
not acquired by that group at the time of the data collection
investigated children from 1st to 3rd grade, and revealed that
(note that this is in line with the accuracy rates of this group).
children responded more accurately to words from dense
What do the present neighborhood density effects tell us
neighborhoods (words with many substitution neighbors)
about lexical access in children? On the one hand, many
than to words from sparse neighborhoods (words with few
researchers have consistently found orthographic effects with
substitution neighbors) or hermits (words with no substitution
beginning readers (Castles, Davis, & Forster, 2003; Perea &
neighbors). In fact, these results entitled the seminal work
Estévez, 2008; Sebastián-Gallés & Parreño, 1995). These
by Laxon et al. (1988): ‘Children find friendly words friendly
results have led to the conclusion that children do develop
too: words with many orthographic neighbors are easier to
preferences for the use of the lexical route, more than
read and spell’. All the effects they showed were based on
grapheme-to-phoneme conversion rules. This is totally in
accuracy analyses, revealing that words from dense
line with connectionist interactive models of reading
neighborhoods and words from sparse neighborhoods
acquisition and development (e.g., McClelland, 1989), and
differed in more than 10% in terms of correct responses.
partially challenges stage models, which assume that there
This difference is enormous when compared to adult data,
is a progression from logographic to alphabetic to
which normally shows differences not higher than 4%. This
orthographic representations in separated stages (Laxon et
difference motivated the current study, since it looked like
al., 1994). On the other hand, some other researchers propose
progressive reading skill acquisition could modulate
that these preferences for using the lexical route are language-
dependent, claiming that young readers from transparent
ORTHOGRAPHIC EFFECTS IN CHILDREN
languages might preferably use grapheme-to-phoneme
Andrews, S. (1992). Frequency and neighborhood effects on lexical
conversion mechanisms (Ziegler & Goswami, 2006). For
access: Lexical similarity or orthographic redundancy? Journal
this last claim to be valid, we should have found no (or little)
of Experimental Psychology: Learning, Memory and Cognition,
neighborhood density effects in Spanish, since it is a
consistent/transparent language with clear correspondence
Andrews, S. (1996). Lexical retrieval and selection processes:
between graphemes and phonemes (note that all the previous
Effects of transposed-letter confusability. Journal of Memory
evidence with children’s N effects was obtained in an opaque
& Language, 35, 775-800.
language such as English). It is clear, however, that this was
Andrews, S. (1997). The role of orthographic similarity in lexical
not the case in the present study: novel readers from a
retrieval: Resolving neighborhood conflicts. Psychonomic
transparent language do show strong N effects.
Bulletin and Review, 4, 439-461.
Further, there are interesting recent data that, together
Balota, D.A., & Chumbley, J.I. (1984). Are lexical decisions a good
with ours, rule out Ziegler and Goswami’s argument (2006).
measure of lexical access? The role of word frequency in the
Lavidor et al., (2006) studied the case of F.M., a man with
neglected decision stage. Journal of Experimental Psychology:
dyslexia, severe phonological problems and great difficulty
Human Perception and Performance, 10, 340-357.
in correctly developing grapheme-to-phoneme conversion
Carreiras, M., Perea, M., & Grainger, J., (1997). Effects of
rules. They studied the N density effects for this man, and
orthographic neighborhood in visual word recognition: cross-
revealed that the N effect was much larger for this man and
task comparisons. Journal of Experimental Psychology:
other dyslexic students, than for a control group (see also
Learning, Memory and Cognition, 23, 857-871.
Snowling, 2000). Lavidor and colleagues concluded that
Castles, A., Davis, C., & Forster, K.I. (2003). Word recognition
“people with dyslexia are relatively advantaged in processing
development in children: Insights from masked-priming. In S.
words that share letter strings with orthographic neighbors,
Kinoshita & S. Lupker (Eds.), Masked priming: State of the
because of a preference for global processing that capitalizes
art (pp. 345-360). Hove, UK: Psychology Press.
on the coarse coding of orthography-phonology mappings”
Chambers, S.M. (1979). Letter and order information in lexical
(p.325; see Harm, McCandliss, & Seidenberg, 2003, for a
access. Journal of Verbal Learning and Verbal Behavior, 18,
similar view). If N effects are generally obtained when a
direct lexical route is used, no grapheme-to-phoneme
Clark, H.H. (1973). The Language-as-fixed-effect fallacy: A critique
conversions seem to be involved in these effects and reader
of language statistics in psychological research. Journal of
groups from both transparent and opaque language show
Verbal Learning and Verbal Behavior, 12, 335-359.
these effects, then there is no reason a priori to discard the
Coltheart, M. (1980). Deep dyslexia: A right hemisphere hypothesis.
lexical route as the mainly adopted route even by very young
In M. Coltheart, K. Patterson, & J.C. Marshall (Eds.), Deep
beginning readers. However, since no measures of grapheme-
Dyslexia (pp. 361-406). London: Routledge & Kegan Paul.
to-phoneme reading were employed, these ideas are just
Coltheart, M. (2000). Deep dyslexia is right-hemisphere reading.
speculation at this stage. Future research in this direction
Brain and Language, 71, 299-309.
should be done in order to clarify the role of grapheme-to-
Coltheart, M., Davelaar, E., Jonasson, J.T., & Besner, D. (1977).
phoneme conversion routines in comparison with direct
Access to the internal lexicon. In S. Dornic, (Ed.), Attention
and performance VI. Hillsdale, NJ: Erlbaum .
In summary, the present experiment provides evidence
Coltheart, M., Rastle, K., Perry, C., Ziegler, J., & Langdon, R.
supporting the early existence of substitution neighborhood
(2001). DRC: A dual-route cascaded model of visual word
density effects in children. Moreover, we have shown that
recognition and reading aloud. Psychological Review, 108, 204-
the pattern of the effects in children is quite similar to the
pattern of the effects in adult readers. Both adults and
Davis, C.J. (1999). The self-organising lexical acquisition and
children «find friendly words friendly». Finally, we have
recognition (SOLAR) model of visual word recognition.
provided evidence in favor of a predominant use of the
Unpublished doctoral dissertation, University of New South
lexical route in beginning readers.
Davis, C.J., & Perea, M. (2005). BuscaPalabras: A program for
deriving orthographic and phonological neighborhood statistics
and other psycholinguistic indices in Spanish. Behavior
Research Methods, 37, 665-671.
Acha, J., & Perea, M. (in press). The effects of length and
Davis, C.J., & Perea, M. (2007). Re(de)fining the orthographic
transposed-letter similarity in lexical decision: Evidence with
neighbourhood: Evidence for the similarity of addition and
beginning, intermediate, and adult readers. British Journal of
deletion neighbours in Spanish. Manuscript submitted for
Andrews, S. (1989). Frequency and neighborhood effects on lexical
Davis, C.J., & Taft, M. (2005). More words in the neighborhood:
access: Activation or search? Journal of Experimental
Interference in lexical decision due to deletion neighbors.
Psychology: Learning, Memory and Cognition, 15, 802-814.
Psychonomic Bulletin & Review, 12, 904-910.
DUÑABEITIA AND VIDAL-ABARCA
Duñabeitia, J.A., Perea, M., & Carreiras, M. (2007). Do transposed-
Lavidor, M., Johnston, R.S., & Snowling, M.J. (2006). When
letter similarity effects occur at a morpheme level? Evidence for
phonology fails: Orthographic neighbourhood effects in
morpho-orthographic decomposition. Cognition, 195, 691-703.
Dyslexia. Brain & Language, 96, 318-329.
Duñabeitia, J.A., Carreiras, M., & Perea, M. (in press). Are coffee
Lavidor, M., & Walsh, V. (2003). A magnetic stimulation examination
and toffee served in a cup? Ortho-phonologically mediated
of orthographic neighbourhood effects in visual word recognition.
associative priming. Quarterly Journal of Experimental Psychology.
Journal of Cognitive Neuroscience, 15, 354-363.
Laxon, V., Coltheart, V., & Keating, C. (1988). Children find
Duñabeitia, J.A., Vidal-Abarca, E., & Izquierdo, C. (2008).
friendly words friendly too: Words with many orthographic
Evaluación del reconocimiento visual de palabras mediante
neighbours are easier to read and spell. British Journal of
tarea de decisión léxica. Manuscript in preparation.
Educational Psychology, 58, 103-119.
Forster, K.I. (1976). Accessing the mental lexicon. In R.J.Wales
Laxon, V., Gallagher, A., & Masterson, J. (2002). Word length and
& E.Walker (Eds.), New approaches to language mechanisms
orthographic neighbourhood size effects in children’s reading.
(pp. 257-287). Amsterdam: North-Holland.
British Journal of Psychology, 93, 269-287.
Forster, K.I., & Forster, J.C. (2003). DMDX: A Windows display
Laxon, V., Masterson, J., & Moran, R. (1994). Are children’s
program with millisecond accuracy. Behavior Research
representations of words distributed? Effects of orthographic
Methods, Instruments, & Computers, 35, 116-124.
neighbourhood size, consistency, and regularity of naming.
Forster, K.I., & Hector, J. (2002). Cascaded versus noncascaded
Language and Cognitive Processes, 9, 1-27.
models of lexical and semantic processing: The turple effect.
McClelland, J. L. (1989). Parallel distributed processing: Implications
Memory & Cognition, 30, 1106-1117.
for cognition and development. In Morris, R. (Ed.), Parallel
Forster, K.I., & Shen, D. (1996). No enemies in the neighborhood:
distributed processing: Implications for psychology and
Absence of inhibitory effects in lexical decision and
neurobiology. (pp. 8-45). New York: Oxford University Press.
categorization. Journal of Experimental Psychology: Learning,
McClelland, J.L., & Rumelhart, D.E. (1981). An interactive activation
Memory, & Cognition, 22, 696-713.
model of context effects in letter perception: Part 1. An account
Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.),
of basic findings, Psychological Review, 88, 375-407.
Cognitive processes in spelling. London: Academic Press.
Paap, K.R., & Johansen, L. S. (1994). The case of the vanishing
Frith, U. (1985). Beneath the surface of developmental dyslexia.
frequency effect: A retest of the verification model. Journal
In K.G. Patterson, J.C. Marshall, & M. Coltheart (Eds.), Surface
of Experimental Psychology:
Human Perception &
dyslexia: Studies of phonological reading (pp. 301-330).
Performance, 20, 1129-1157.
London: Routledge and Kegan Paul.
Paap, K.R., Newsome, S.L., McDonald, J.E., & Schvaneveldt,
Frost R., Kugler T., Deutsch A., & Forster K.I. (2005). Orthographic
R.W. (1982). An activation-verification model for letter and
structure versus morphological structure: Principles of lexical
word recognition: The word-superiority effect. Psychological
organization in a given language. Journal of Experimental
Review, 89, 573–594.
Psychology: Learning, Memory and Cognition, 31, 1293-326.
Page, M. (2000). Connectionist modelling in psychology: A localist
Gómez, P., Ratcliff, R., & Perea, M. (2007). A model of letter
manifesto. Behavioural and Brain Sciences, 23, 443-67.
position coding: The overlap model. Manuscript submitted for
Perea, M. (1998). Orthographic neighbours are not all equal:
Evidence using an identification technique. Language and
Grainger, J., Granier, J.P., Farioli, F., Van Assche, E., & van Heuven,
Cognitive Processes, 13, 77-90.
W. (2006). Letter position information and printed word
Perea, M., & Algarabel, S. (1999). Puntuaciones típicas y potencia
perception: The relative-position priming constraint. Journal
estadística con diferentes procedimientos de análisis de los tiempos
of Experimental Psychology: Human Perception and
de reacción: un estudio de simulación. Psicológica, 20, 211-226.
Performance, 32, 865-884.
Perea, M., & Estévez, A. (2008). Transposed-letter similarity effects
Grainger, J., & Jacobs, A.M. (1996). Orthographic processing in
in naming pseudowords: Evidence from children and adults.
visual word recognition: A multiple read-out model.
European Journal of Cognitive Psychology, 20, 33-46.
Psychological Review, 103, 518-565.
Perea, M., & Fraga, I. (2006). Transposed-letter and laterality
Grainger, J., & Seguí, J. (1990). Neighborhood frequency effects
effects in lexical decision. Brain and Language, 97, 102-109.
in visual word recognition: A comparison of lexical decision
Perea, M., & Lupker, S.J. (2003). Transposed-letter confusability
and masked identification latencies. Perception and
effects in masked form priming. In S. Kinoshita & S.J. Lupker
Psychophysics, 47, 191-198.
(Eds.), Masked priming: State of the art (pp. 97-120). Hove,
Harm, M.W., McCandliss, B.D., & Seidenberg, M.S., (2003).
UK: Psychology Press.
Modelling the successes and failures of interventions for
Pollatsek, A., Hyönä, J., & Bertram, R. (2000). The role of
disabled readers. Scientific Studies of Reading, 7, 155-182.
morphological constituents in reading Finnish compound words.
Holcomb, P.J., Grainger, J., & O’Rourke, T. (2002). An
Journal of Experimental Psychology: Human Perception and
electrophysiological study of the effects of orthographic
Performance, 26, 820–833.
neighborhood size on printed word perception. Journal of
Pollatsek, A., Perea, M., & Binder, K. (1999). The effects of
Cognitive Neuroscience, 14, 938-950.
neighborhood size in reading and lexical decision. Journal of
ORTHOGRAPHIC EFFECTS IN CHILDREN
Experimental Psychology: Human Perception and Performance,
Siakaluk, P.D., Sears, C.R., & Lupker, S.J. (2002). Orthographic
neighborhood effects in lexical decision: The effects of
Ratcliff, R. (1993). Methods for dealing with reaction time outliers.
nonword orthographic neighborhood size. Journal of
Psychological Bulletin, 114, 510-532.
Experimental Psychology: Human Perception and Performance,
Rubinstein, O., Henik, A., & Dronkers, N.F. (2001, September 4-
7). Is the right hemisphere capable of reading and
Snodgrass, J.G., & Mintzer, M. (1993). Neighborhood effects in
understanding a word? Presented in ESCOP, Edinburgh, UK.
visual word recognition: Facilitatory or inhibitory? Memory &
Rumelhart, D.E., McClelland, J.L., and the PDP research group. (1986).
Cognition, 21, 247-266.
Parallel distributed processing: Explorations in the microstructure
Snowling M.J. (2000). Language and literacy skills: who is at risk
of cognition. Volume I. Cambridge, MA: MIT Press.
and why? In D.V.M. Bishop & L.B. Leonard, (Eds.). Speech
Sears, C.R., Hino, Y., & Lupker, S.J. (1995). Neighborhood
and language impairments in children: Causes, characteristics,
frequency and neighborhood size effects in visual word
intervention and outcome (pp. 245-260). Hove, UK: Psychology
recognition. Journal of Experimental Psychology: Human
Perception & Performance, 21, 876-900.
Spieler, D.H., & Balota, D.A. (2000). Factors influencing word naming
Sebastián-Gallés, N., Cuetos, F. / Carreiras, M. / Martí, M. A.
in younger and older adults. Psychology and Aging, 15, 225-231.
(2000) Lexesp. Léxico informatizado del español. Barcelona:
Whitney, C. (2001). How the brain encodes the order of letters in
Publicacions i Edicions UB
a printed word: The SERIOL model and selective literature
Sebastián-Gallés, N., & Parreño, A. (1995). The development of
review. Psychonomic Bulletin and Review, 8, 221-243.
analogical reading in Spanish. Reading and Writing, 7, 23-
Ziegler, J.C., & Goswami, U. (2006). Becoming literate in different
languages: similar problems, different solutions. Developmental
Shelton, J.R., & Martin, R.C. (1992). How semantic is automatic
Science, 9, 429-453.
semantic priming? Journal of Experimental Psychology:
Learning, Memory, and Cognition, 18, 1191-1210.
Shillcock, R., Ellison, T. M., & Monaghan, P. (2000). Eye-fixation
Received April, 20, 2007
behavior, lexical storage and visual word recognition in a split
Revision received June, 26, 2007
processing model. Psychological Review, 107, 824-851.
Accepted September, 24, 2007
The 44 words used in the present study (in parenthesis their neighborhood density): ahí(4), ajo(9), bajo(8), caliente(2),
centro(1), charco(4), ciervo(6), deuda(0), diez(1), estado(1), fiel(5), fresca(3), islote(0), jefe(1), largo(3), lío(8),
llanura(0), luego(3), lugar(5), menta(8), miedo(0), olfato(0), pajar(6), paz(6), persona(0), planeta(0), proceso(1),
puerta(6), reja(7), riesgo(0), roce(3), rural(1), siesta(1), sombra(1), suela(4), texto(1), treinta(0), triste(1), valiente(2),
ventaja(1), vía(11), virus(0), viuda(1), voto(11).