Mojibake – The rehearsal of word fragments in verbal recall
of verbal rehearsal usually assume that whole words are being
rehearsed. However, words consist of letter sequences, or syllables, or
word onset-vowel-coda, amongst many other conceptualizations of word
structure. A more general term is the ‘grain size’ of word units (Ziegler and Goswami, 2005).
In the current study, a new method measured the quantitative percentage
of correctly remembered word structure. The amount of letters in the
correct letter sequence as per cent of word length was calculated,
disregarding missing or added letters. A forced rehearsal was tested by
repeating each memory list four times. We tested low frequency (LF)
English words versus geographical (UK) town names to control for
content. We also tested unfamiliar international (INT) non-words and
names of international (INT) European towns to control for familiarity.
An immediate versus distributed repetition was tested with a
between-subject design. Participants responded with word fragments in
their written recall especially when they had to remember unfamiliar
words. While memory of whole words was sensitive to content,
presentation distribution and individual sex and language differences,
recall of word fragments was not. There was no trade-off between memory
of word fragments with whole word recall during the repetition, instead
also word fragments significantly increased. Moreover, while whole word
responses correlated with each other during repetition, and word
fragment responses correlated with each other during repetition, these
two types of word recall responses were not correlated with each other.
Thus there may be a lower layer consisting of free, sparse word
fragments and an upper layer that consists of language-specific,
orthographically and semantically constrained words.
is one of the most interesting phenomena because it captures the
transition from the first strenuous effort at solving a task to an
automatized and much more effortless process (Logan, 1990; Fecteau and Munoz, 2003).
In verbal memory, rehearsal develops at about 7 years as indicated by
the onset of the phonological similarity effect at this age (Jarrold and Tam, 2011,
p. 186), yet these authors hold that the onset of verbal rehearsal in
general may nevertheless be gradual rather than discrete. Indeed, in the
development of reading, repetition was shown to be of major importance
already at a young age (Horst et al., 2011; Horst, 2013).
The effect of repetition is also extensively exploited in supervised
neural networks where in each repetition an error feedback signal is
considered in order to optimize learning (McLeod et al., 1998).
The adaptation of the neural structure often takes many sweeps. To take
time to memorize to perfection by rote learning was already measured in
1885 (Ebbinghaus, 1964).
Ebbinghaus meticulously recorded the time it took him to learn
non-sense syllables perfectly by heart and found that on each
repetition, he needed less time to achieve the same performance level.
Furthermore, a neural network simulation using the original Ebbinghaus
stimuli as input showed that the network learned better and more
accurately without transformational (conceptual) hidden nodes, but
produced the same output as input with a direct mapping
approach. Networks always needed 200 sweeps independently whether a
graphic or a phonological code was used, or homogeneous or mixed lists
had to be learned – it just queued the stimuli into a sequence for
output (Lange-Küttner, 2011; see also Mitchell and Zipser, 2003).
early Ebbinghaus experiment showed that we do not necessarily need
rehearse just whole words. The current study investigates whether
rehearsal in a verbal recall task may actually involve word fragments.
This hypothesis is backed up by recent work that shows that word
structure is relevant for reading (Lange-Küttner, 2005) as well as for word memory (Lange-Küttner and Krappmann, 2011). In neural networks and reading research, usually the word onset, vowel and coda (Plaut et al., 1996) or the ‘grain size’ of units (Ziegler and Goswami, 2005)
are distinguished as building blocks of a word. In memory research,
participants could visually recognize word fragments that they had seen
in a presentation – even if they were unable to complete the word
fragment into a whole word (Challis and Sidhu, 1993; Nyberg et al., 1994; Cleary and Greene, 2000). Meaningful fragments (Cleary, 2002) and more frequent fragments (Cleary and Greene, 2001)
were easier to recognize. Already 5-year-old British children who are
beginning to read are able to recognize word fragments such as ‘bzn’ for
the word ‘basin’ and they can even distinguish it from another fragment
where instead of the phonetic cue ‘z’ for the word ‘basin’ a control
cue ‘f’ is used (Rack et al., 1994).
In the current study, only whole words and pseudo-words varying in
familiarity and content were presented. Instead, we analyzed whether
participants generated word fragments when writing down their responses
in the recall phase of the word memory experiment. We used a new method
that measured the percentage of correctly remembered word structure.
From participants’ written recall of words, we scored not only the
correctly sequenced words, but also the amount of letters in a correct
sequence in fragments of a proper word. We disregarded missing or added
letters, and just computed the number of correctly recalled letters as
the percentage of the actual word length, because the word length effect
is one of the most robust effects in word memory (Baddeley et al., 1975).
specifically, we hypothesized that like young infants who gradually
learn the correct pronunciation of a word in their spoken word
production and simultaneously drop their approximations and inventions (Dromi, 1987),
the young adults in the current study would be able to gradually write
the correct orthography of a memorized word in their written word
production in a word recall task when trying several times. All
participants repeated the recall of the word lists four times, because
we know that rehearsal and repetition enhances word memory as such as
well as the length of a word that can be remembered (Samuels et al., 1979).
We were interested whether verbal recall would improve more when the
words were immediately repeated in the next three blocks, or whether a
less forceful rehearsal with a randomly distributed encounter of each
word list would facilitate word recall more. The distinction of massed
versus distributed practice in verbal learning usually refers to the
length of the inter-stimulus interval (ISI). Underwood (1961)
claimed that a long ISI allows time for the successive extinction of
errors, while a short ISI would suppress errors rather than extinguish
thus each block had the same length. However, we did vary the sequence
of the blocks in order to test massed versus distributed practice. One
group of participants experienced each list four times in immediate
succession in a kind of forced rehearsal. The other group of
participants also experienced each word list four times, but the
repeated word lists were presented in a mixed sequence randomized by the
computer program for more incidental learning. Our prediction was that
immediate repetition of a word list would support verbal recall more
than a randomly distributed repetition. We assumed that an immediate
repetition would also have a stronger effect because it resembles the
spontaneous rehearsal of children and adults when they try to keep words
in the mind for fast and safe retrieval.
Because we tested mainly young people with different ethnic backgrounds
and from many countries who often spoke more languages than just
English, we also monitored the content of the words. We tested names of
British and international (INT) European towns (no capitals and
controlled for town size). We expected that the UK towns would be easier
to remember than the INT towns because of a geocentric memory bias (Baddeley, 1999,
pp. 158–159). Furthermore, we tested low frequency (LF) English words
against INT non-words to control familiarity. INT non-words were
previously used for a cross-cultural comparison of word reading in young
adults (Paulesu et al., 2000) and vocabulary learning in children (Morra and Camba, 2009) in order to avoid that non-words would vary in familiarity like existing lexical items (Treiman et al., 1990).
We created the INT non-words by translating the English LF words into
German, Danish, French, Italian, and Spanish. We then randomly took two
or three word fragments (depending on the length of the original word)
and combined them into a new word which contained legal letter sequences
from the foreign words.
we designed a word recall task that controlled for repetition intensity,
content and familiarity. The theoretically relevant idea for this
investigation into visual word memory is to evaluate the memory
fragments that the young people recalled instead of only the absolutely
correct whole words that were remembered. In this way, we may be able to
discover whether visual word memory rehearsal also involves word
fragments, and whether these remembered fragments are a gradual
approximation toward memorizing whole words. Similar response
evaluations that distinguished between partially correct and completely
correct responses were conducted in research with spoken stimuli and
spoken responses (Storkel et al., 2006) which will allow comparison in the Discussion.
Materials and Methods
= 43 bilinguals (26 females). All participants were students of the
London Metropolitan University, City Campus. The mean age was 27 years
(SD = 9 years, range 18–55 years). Monolinguals were native
English-speakers with British or US nationality. Nationalities of the
bilinguals varied widely, with 27 different nationalities.
The LF words were all nouns with relatively different translations in
German, Danish, French, Italian, or Spanish. We did not use words like
‘monarchy’ that would have been nearly the same in all the translations.
Word fragments used for the creation of the new INT words are set in
bold in Table Table11.
as far as possible because also the size of the towns in terms of
number of inhabitants was controlled. No names of capital towns were
used. Combined letters such as ‘st’ or ‘aa’ or ‘nn’ or ‘ei’ were counted
as one letter when spoken as one sound. Consonant clusters are used as
one sound in experimental studies (e.g., Page et al., 2006,
p. 726) and their letter count varied on average by one or two letters
per word. Although consonant clusters, such as ‘st,’ are not listed in
the IPA phonetic alphabet, the phonetic voiceless alveolar sibilant
consonant ∫can be joined by a tie bar if for instance merged with
another sound like in ‘st’ or ‘sch’ in another language (International Phonetic Association, 2005).
Averaged values give information about the overall ease of
pronunciation of a word because difficult phoneme transitions can be
ameliorated by easier phoneme transition (Coleman and Pierrehumbert, 1997).
These PTVs are more commonly used in studies where words need to be
articulated as part of the experimental design in order to control for
the ease to pronounce a word. Ease of word pronunciation according to
PTV makes overt word repetition easier and has an interactive
relationship with vocabulary size in children and adults (Edwards et al., 2004; Munson et al., 2005). It also facilitates repetition level but not repetition rate in neural networks (Gupta and Tisdale, 2009).
the current study investigated visual word memory, that is,
participants saw the words and wrote down the words without a word being
said. Thus the PTV was not an experimental design factor. We also did
not translate the memory items into a Klattese transcription, but
entered them as correctly spelled words – as the participants
encountered them in the experiment – into a Phonotactic Probability
Calculator that operates on the basis of an English language word data
base (Kucera and Francis, 1967). The resulting PTV was related to word length as shorter words had lower values which conforms with earlier research (Bailey and Hahn, 2001).
Furthermore, PTVs were not related to familiarity as the international
INT words had relatively similar PTVs to the more familiar words. Also
this result is in agreement with Bailey and Hahn (2001)
who emphasized that PTVs can vary more drastically between native
English words than in comparison to INT words. In the current study,
similarities may have occurred because the INT non-words and place names
were all from West European areas.
when comparing the word lists, the PTVs were relatively homogeneous. The
average PTV of the four memory lists in Word Lists 1 was M = 25.5 (range 22–28) and in Word Lists 2 it was M
= 28.7 (range 27–32). All memory lists in Word Lists 1 were tested
before those in Word lists 2. Accordingly, block sequence was separately
permutated for Word Lists 1 and Word Lists 2. The computer programming
software Experimental Run Time System (ERTS; Beringer, 1994)
was used to present the word lists and instruct the participants. One
word was presented at a time in a randomized sequence on a DOS computer
with a 15 inch screen. Each word was presented in Times Large 12 font in
white on a black background for 1000 ms, with an ISI of 500 ms. The
presentation of the words occurred in blocks of eight words (see Table Table22).
Because each word list was repeated four times, each set had 16 uniform
word list presentations. In the rehearsal condition, each word list
type was immediately repeated. In contrast, in the incidental learning
condition, the sequence of the four times repeated blocks A, B, C, and D
was randomly and completely permutated by the ERTS within each set,
rather than at fixed intervals (Page et al., 2013). For instance, memory word list A1-4 could be repeated at any place in the random sequence of 16 blocks (e.g., A1, B1, D1, C1, A2, A3, B2 …) and the maximum possible space between repetitions of block type A1 and A2 was about 12 blocks if the first block was repeated only at the end of the set (e.g., A1, C1, D1, B1, D2, C2, B2, C3, D3, B3, B4, C4, A2, D4, A3, A4).
were tested individually in a quiet computer laboratory. The experiment
was vetted and approved by the Departmental Ethics Committee. Before
the start of the experiment, participants were provided with a Consent
form which they signed. Afterward, they received a Debrief form for
informative details about the nature of the study. They were randomly
assigned to one of two experimental conditions – condition 1 (immediate
block repetition, forced rehearsal) or condition 2 (program generated
permutated block sequence, incidental learning condition). Participants
were provided with paper notepads to write down their responses. They
turned over a sheet after each word list.
were given in written form on the computer display. Participants were
informed that some words made sense, while others would not, and that
each word list would be repeated four times during the course of the
experiment. Their task was to recall as many words as they could
remember. At the end of each word list presentation they were asked to
write down the words on paper in any order in which they came to mind
(free recall). There was no delay after the presentation of each word
list and recall time was not constrained. Participants pressed the space
bar to initiate the next word list (self-paced block transitions).
responses were scored twice. Firstly, we scored correctly memorized and
orthographically correctly written whole words. Accuracy was computed
per block in per cent correct. We also scored remembered words that were
recognizably part of the memory list but consisted of word fragments
with only some letters in the correct sequence. We disregarded missing
(omissions) or added (intrusions) letters. For instance, one word in the
UK places list was ‘Salisbury.’ In the response word ‘Sailsbry’ (which
has half a word with a different meaning denoting the sails of a boat),
the letter ‘i’ is in the wrong place and the letter ‘u’ is missing, but
all other letters are in a correct sequence. Thus, the participant
scored 7 letters out of 9 correct, and received a score of 7/9 = 77.8%
accuracy for this word. In another scoring example, a participant wrote
‘Sainsbury,’ that is, also this participant made a semantic mistake, but
for the whole word. The participant remembered a similarly written word
that denoted a British supermarket instead of a British town. In this
word, the letter ‘l’ is missing and the letter ‘n’ is a wrong letter,
but all other letters are in the correct sequence, 8/9 = 88.9% correct
(% correct per word). These examples show that the meaning of the
associated word may only have been a memory trigger as the semantic
association could be quite remote to the actual stimulus, while the
important feature is the orthographic similarity with the target word.
The results from the accurate and the more lenient scoring were averaged
per word type list, respectively, across memory list 1 and 2.
because the lenient scoring yielded higher accuracy scores, we computed
a stricter score for correct whole words which was then subtracted from
the values that were obtained with the lenient scoring. The resulting
scores were the pure values for just the ‘nearly correct words’ (word
fragments) which we then compared with the whole word score. The
comparison allowed to test whether the effect of repetition (rehearsal)
relates not only to whole words but also to word fragments. If there is a
gradual approximation during rehearsal toward the correct whole word,
we expected that word fragments should decrease during
rehearsal/repetition and would correlate with the whole word score in
the subsequent block.
first analysis compares the two scoring methods. Recall scores were
analyzed with analyses of variance with repeated measures. When the
Mauchley’s Test of Sphericity was significant, the degrees of freedom
were adjusted according to Huynh-Feldt. In a second analysis of variance
thereafter, a fragments-only score was analyzed.
conducted a 2 (Words/Places) × 2 (Familiarity) × 4 (Repetition) × 2
(Scoring Method) × 2 (Training) analysis of variance with repeated
measures on the first four factors, and type of training as a
between-subject factor. Differences due to age were partialled out using
the variable ‘age in years’ as covariate. In an initial analysis, we
also included the variables sex and language of the participants as
between-subject factors. However, the inclusion of these individual
difference factors made the analysis of variance very complex. Like Logie (2011,
p. 243) predicted, the main experimental results did not change when
the individual difference variables were omitted. In short, men showed a
memory advantage for INT places. Bilinguals profited somewhat more from
immediate repetition, while monolinguals benefited from incidental
learning especially when words were unfamiliar. Because these results of
individual differences did not substantially contribute to the
hypothesis, the statistical details of this initial analysis are not
and are not quoted again in the text. The main effect of training type
(immediate vs. distributed repetition of blocks) was not significant as a
between-subject factor showing that memory performance in general did
not vary in the two training groups. The main effects of scoring,
familiarity and repetition were all highly significant, ps
< 0.001. As expected, participants showed better word memory when
also correct letter sequences in word fragments were scored (M = 57.2%) rather than just whole words (M = 49.4%). Participants remembered the familiar LF English words and UK places (M = 67.8%) better than the unfamiliar words, that is the INT non-words and INT places (M
= 38.8%). Furthermore, mere repetition nearly doubled word memory
accuracy which supports the hypothesis that prescribed rehearsal is an
efficient facilitator (Block 1 M = 38.3%, Block 2 M = 53.0%, Block 3 M = 60.1%, Block 4 M = 61.8%).
= 80, for Scoring Method (left) and Composite Scores (right) Analyses
of variance with repeated measures for Repetition (four times),
Familiarity (low/high) and Content (words vs. geographical places).
= 32.0%). As would be expected, the more challenging INT words
benefited considerably from the additional letter sequence scoring
compared to the whole word scoring, that is an increase of above 10%
occurred (INT non-words M = 37.0%/M = 26.9%; INT places M = 51.1%/M = 40.2%). In contrast, the UK places benefited only 6.2% (M = 70.8%/M = 64.6%) and the LF English words just 3.8% (LF English words M = 69.8%/M = 66.0%). Thus, the more difficult a word was to remember the more it benefited from the additional letter sequence scoring.
(two-tailed) comparing the two rehearsal conditions per block showed
that the immediate repetition advantage only gradually emerged for the
INT words, that is, for INT places [Block 1 t(78) = 1.93, p = 0.058; Block 2 t(78) = 1.56, p = 0.123; Block 3 t(78) = 1.43; p = 0.157; Block 4 t(78) = 2.04, p = 0.044] and especially for the INT non-words [Block 1 t(78) = 0.543, p = 0.058; Block 2 t(78) = 1.44, p = 0.153; Block 3 t(78) = 2.74; p = 0.008; Block 4 t(78) = 3.03, p = 0.003].
Immediate repetition is denoted by the solid lines, the permutated
memory block condition is denoted by the broken line. Immediate
repetition is significantly more efficient for word memory recall than
incidental encounters (permutated ...
(two-tailed) showed that memory for geographic place names was better
than for words. This difference stayed significant throughout the
experiment, ts (79) > -4.03, ps < 0.001, see Figure Figure22. Correlations between memory for place names and words increased when only whole words were scored (Block 1 r = 0.51, Block 2 r = 0.68, Block 3 r = 0.79, Block 4 r = 0.79), and also when letter sequences were scored in addition (Block 1 r = 0.50, Block 2 r = 0.72, Block 3 r = 0.84, Block 4 r = 0.82), ps
< 0.001. These results suggest that the content of the words became
less important for memory performance during the experiment because the
shared variance between the two types of memory lists increased. Figure Figure22
shows that indeed there is a subtle narrowing of the gap between place
names and words during practice which is slightly more pronounced when
scoring letter sequences (difference between initial and final gap =
2.51%) than when scoring whole words (difference between initial and
final gap = 2.19%).
Memory for geographic place names was better than for LF words and
non-words. This difference was somewhat less pronounced for completely
correctly spelled words (pastel colored lines). It stayed significant
throughout the experiment, ...
the interaction of scoring by repetition with familiarity showed that
participants particularly benefited from the scoring method which
appreciated letter sequences when they recalled unfamiliar words as they
were significantly more likely to recall unfamiliar than familiar words
as word fragments, see Figure Figure33.
It was significantly more likely that unfamiliar words were recalled as
word fragments (letter sequences) than familiar words. Means are
controlled for age. Bars denote the SE.
second analysis compared the whole word score with the partial word
fragment score as described before in the Methods section. Again, recall
scores were analyzed with analyses of variance with repeated measures.
We conducted again a 2 (Words/Places) × 2 (Familiarity) × 4 (Repetition)
× 2 (Scores) × by 2 (Training) analysis of variance with repeated
measures on the first four factors, type of training as between-subject
factor and age in years as covariate. Also in this second analysis, the
statistical effects are listed in Table Table33, on the right hand side. Most of the statistical effects are the same, however, there are some important differences.
effect size of the scoring effect more than doubled. This occurred
because there were on average significantly fewer word fragments (M = 7.7%) than totally correct words (M
= 57.2%). The scoring effect interacted with familiarity and training.
This interaction was further explored with two MANCOVAs (controlled for
age) separately for whole words and fragments, respectively. The type of
repetition mattered for whole words, F(1, 80) = 10.25, p = 0.002, h2 = 0.12. When familiar words were encountered it made little difference whether they were immediately (M = 69.8%) or incidentally (M
= 70.8%) repeated. In contrast, when unfamiliar words were encountered,
these were better remembered when repeated in immediate succession (M = 47.5%) than when encountered incidentally (M = 40.5%), t(78) = 2.26, p
= 0.027 (two-tailed). The type of training did not matter for word
fragments, though: In both memory training conditions, unfamiliar words
were more likely to be recalled as a fragment (immediate M = 10.8%, incidental M = 10.1%) than familiar words (immediate M = 5.9%, incidental M = 4.2%).
for the rehearsal hypothesis, the scoring effect interacted with
repetition, with a relatively large effect size of η2 = 0.44,
and this effect did not interact with the timing of the blocks. The
amount of orthographically correct recalled words increased with
repetition by 24.4% (Block 1 M = 41.6%, Block 2 M = 56.9%, Block 3 M = 64.1%, Block 4 M = 66.0%), while the word approximations increased by 1.8% (Block 1 M = 6.6%, Block 2 M = 7.9%, Block 3 M = 8.0%, Block 4 M = 8.4%). There was no decrease in word fragments.
there was the possibility that the word fragments did not increase as
much because they were feeding into the increase of correctly spelled
words. To investigate this question, we computed the four correlations
between the four repeated blocks of the whole words and the word
fragments, respectively, and the three correlations of word fragments
with the subsequent block of whole words. We adapted the level of
significance to 0.05/11 correlations = p < 0.004. The correlations in Table Table44
show the same correlational pattern for the total sample as for all
four sub-samples. The repeated recall of correctly spelled whole words
correlated highly and significantly with each other, and likewise, word
fragments where only some letters were in the correct sequence
correlated significantly with each other from one block to the next,
although at a somewhat lower level. However, the recalled word fragments
showed not a single significant correlation with correct whole word
We do not doubt these findings, but we do doubt that the phonological
loop and (sub-vocal) articulation are the only relevant systems of word
memory. Page et al. (2006,
p. 732), for instance, write that when access to the loop would be
blocked by concurrent articulation, participants would need to fall back
‘on a largely unrehearsable visual store.’ Importantly for the current
study, Page and Norris (2009)
assume that the repetition and rehearsal of a word actually builds a
long-lasting long-term memory (LTM) representation, but of a
phonological word-form in the mental lexicon. However, Darling and Havelka (2010) and Darling et al. (2012)
suggested instead that visual and verbal information of words are bound
together in the multi-sensory episodic memory system which is
integrated into the working memory system (Logie, 2011; Baddeley, 2012).
Likewise, in many word recognition models in reading research,
grapheme–phoneme correspondences are assumed to be made when reading
aloud (Coltheart, 2012).
the current study, participants were writing down responses in a free
word recall task, and hence they had to resort to visual orthographic
patterns that they saw before. They saw them several times which gave
them the opportunity to improve their memory performance. However,
participants did not only reproduce the actual words from the memory
list, but also wrote down partially correct words. These sublexical word
structures were found to occur also in spoken responses (Storkel et al., 2006). They are also common in children, for instance, Treiman (1993)
showed that first graders’ correct spellings increased within 1 year
from 888 to 1,989 correct spellings (124%), but so did the wrong
spellings from 1,135 to 1,605 wrong spellings (41.4%). Thus, both
accurately spelled words and words with wrong spellings increased,
albeit at different rates. Also in the current experiment with young
adults, however, written word fragments increased during repetition and
this showed no trade-off with correct words. The word fragments were
learned insofar as scores were correlated with each other during
repeated blocks, but not with correct words. This kind of error learning during repetition also occurred in a serial recall task using just letters (Couture et al., 2008) and in a word memory task (Storkel et al., 2006).
In short, the current study makes a case that word fragment learning
showed the two sides of rehearsal and repetition: not just accurate
responses, also the probability of giving a wrong response increases
with the number of prior occurrences of that response.
Visual Orthographic Patterns of Letters
came to the conclusion that there are no brain areas which are
originally reserved for reading words. In development, brain areas with
other primary functions such as syntactic processing when reading
sentences, face recognition in the case of visual complex pattern
recognition when identifying words, or the lexicon for spoken words when
matching phonological forms, are recruited for reading print. The most
direct way to encode in a visual word memory task where words are read
from the screen and written down during recall would be visual mapping (Lange-Küttner, 2014) or visual bootstrapping (Darling et al., 2012).
investigated the effects of repetition in the two modalities. Adults’
learning effects during repetition were locked into one modality without
any transfer in the case of letters and pictures. However, when words
were used, transfer occurred in the visual-then-auditory condition, but
not in the auditory-then-visual condition: A sound was associated with a
visual word seen before, but a visual word was not associated with a
spoken word heard before. Hence, we may be more likely to enliven the
‘graphic imagery’ of a written word with a sound than to think about how
a word is written after hearing it. This seems to indicate that we do
not have much visual imagery for written words.
children, there was a clear culturally and educationally shaped
preference of children to recruit one modality only for reading, either
visual or auditory word memory (Lange-Küttner and Krappmann, 2011). This selectivity in memory has been emphasized since quite some time (Cowan, 1995).
In fact, when neural networks were run, double-modality word input and
double encoding was most beneficial for immediate word reproduction, but
only one working memory system was necessary to integrate a letter
sequence for word learning to occur (Lange-Küttner and Krappmann, 2011).
Boys seem to have a preference for the visual modality which includes
perception of the fine visual detail of the orthographic letter patterns
(Mohamed et al., 2011; Huestegge et al., 2012).
These orthographic patterns are assumed to be stored in the brain in
the ‘visual word form system’ and can be evoked by writing (Dehaene et al., 2010).
The letters in the orthographic pattern allow a much more precise
notation of sounds than is apparent in the sonograms of spoken language
where not only individual words but also individual sounds present a
segmentation problem (Whitney, 1998, p. 142; Lange-Küttner et al., 2013).
The digital transformation of naturally spoken speech into written
words still represents a major challenge for typing software. Likewise,
in school children the best predictor for writing inner speech into a
fluent text – besides writing speed – is word spelling accuracy (Connelly et al., 2012).
Thus, one would suppose that also visual orthographic patterns of
letters are important for verbal memory and can be rehearsed.
found indeed that in the written responses of our participants, mere
repetition nearly doubled accuracy which clearly supports the hypothesis
that prescribed rehearsal is an efficient facilitator for word memory.
For familiar words it made little difference whether they were
immediately or incidentally repeated, while the unfamiliar INT words
were better remembered when repeated in immediate succession. The
unfamiliar INT words were also more likely to be recalled as a fragment
than familiar words. This suggests that unfamiliar words with foreign
spellings benefit from immediate rehearsal that builds up a visual
orthographic template in LTM within a relatively short time. Writing an
unfamiliar word correctly, however, is a fragile process which was not
helped by immediate repetition. In the following section we discuss why
this may be the case.
Error Learning in Word Memory during Repetitions
paid particular attention to the orthography that the participants
produced when recalling the word lists and writing down their responses.
When words were not spelled correctly yet still identifiable as memory
of the correct word, we scored the letters in the right sequence as per
cent of the actual word length. We wanted to know whether these letter
sequence word fragments would develop into proper whole words if
rehearsed several times. We predicted that if this would be the case,
word fragments should decrease during the repetition, while the whole
word score should increase. This hypothesis was partly confirmed. One
the one hand, it was true that rehearsal in the repeated memory blocks
produced a higher whole word memory score, on the other hand, word
fragments did not decrease, but increased too. Thus, the expected
trade-off between word fragments and whole words did not occur. Instead,
also the word fragments increased with repetition, albeit by a smaller
amount, but then there were also fewer word fragments than whole words
in participants’ response sheets. Word fragments were more often
produced in response to unfamiliar words, e.g., in response to the INT
non-words with legal letter combinations from other languages and INT
geographical places also following non-English language spelling rules.
confirmed results from a developmental study with 8- to 10-year-old
children showing that non-words created from the native language were
easier to learn than non-words created from a non-native language (Morra and Camba, 2009).
The increase of word fragments suggests that during the experiment,
participants kept trying to cobble together letters into word patterns
that resembled the visual input word to some degree, not unlike the
5-year-old reading beginners adept in distinguishing visual word
fragments (Rack et al., 1994).
Also in this study, correct recall of a letter in the right place
showed the same learning curve during repetition as erroneously recalled
letters, that is error learning occurred. Interestingly, in the study
of Couture et al. (2008)
the error learning during repetition occurred only when data of real
people were analyzed, but not when simulated data were used which
yielded an increase in correct answers while wrong responses stayed at
The auditory format enabled the authors to control the stimuli for
phoneme transition difficulty (ease of pronunciation) and lexical
neighborhoods (number of similar words), two factors which impact on
non-word learning quite independently of each other (Bailey and Hahn, 2001). Storkel et al. (2006)
showed that scores of both completely correct words and partially
correct sublexical word units increased during repetition. This
repetition effect did neither interact with lexical neighborhood density
nor with phonotactic probability of the words. Correct words increased
at a steady rate throughout seven repetitions, while partially correct
words leveled off after four exposures. However, no statistical
comparison was made which would have shown whether this difference would
have amounted to a significant interaction that denoted a trade-off
between partially correct words and complete words. Thus, in this study
it remains unclear whether adults could transform a spoken word
approximation into a proper word during repetition. To our knowledge,
only two studies so far showed that errors were actually decreasing
during repetition. One study used 10-item digit sequences from 0 to 9 in
an immediate serial recall task (Cumming et al., 2003).
Importantly, errors were omission mistakes where participants would
initially fill in blanks, but during repetition became able to fill the
gaps. The same effect of repetition was found when letter sequences were
used (Couture and Tremblay, 2006,
Experiment 3). However, this was not the case in 5–6 years old children
who improved with repetition, but not by supplementing missing
information in serial positions (Mosse and Jarrold, 2010).
This may have been the case because at this age, children are not yet
fluent readers, and when they spell words with letters, commission
errors are more frequent than omission errors or reversals in the letter
sequence (Treiman, 1993).
Nevertheless, the 5–6 years old children’s learning during repetition
correlated significantly with learning non-words, but not with regular
word learning. Hence, one conclusion could be that the input repetition
effect seems to transform novel information into familiar information
that can potentially be incorporated into a systematic database (Plunkett and Marchman, 1990).
the current study, it is very likely that the correctly written down
whole words were rehearsed via inner speech which speaks to a
straightforward involvement of the lexicon and semantic LTM. Also in the
Storkel et al. (2006) study, memory for complete spoken words was determined by lexical neighborhood density only.
in the current study it is less likely that also the partly correct
written down word fragments were processed via lexical access because
they were immune to content and presentation distribution effects. The
amount of word fragments occurred also independently of individual
differences with regards to sex and language. We must assume that when
word fragments were written down as a response that this visual
orthographic pattern was remembered from the presentation. A fragmented
visual registering of the word input may be responsible for partial
recall because inserting a delay before a recall test which could have
been used for enhanced recovery did not make any learning difference (Oberauer and Meyer, 2009). The repeated learning would then serve as a kind of sensory visual learning (Mortensen and Nachtigall, 2000; Blum and Yonelinas, 2001) until an accurate word form has been registered that can be associated with some meaning. Also in the Storkel et al. (2006)
study with spoken word stimuli, the partially correct words were not
lexically retrieved, but instead phoneme transitions of the words were
important. Hence, one could conclude that learning of novel unfamiliar
words can begin on a very raw sensory level, for spoken words with
acoustic sounds and for written words with graphemes.
result of different processes for complete vs. partial word memory
during rehearsal and repetition is further underpinned by the finding
that the increase in complete words and the increase in word fragments
occurred independently of each other, as we could not find significant
correlations between them. Word fragments in the repetition were highly
and significantly correlated with each other in the total sample, with r
values between 0.69 and 0.75. This was somewhat lower than for whole
words which correlated very highly between 0.83 and 0.87 in the
repetitions. This correlational pattern could be replicated with a
split-file method, with r values between 0.59 and 0.83 for word fragments and r
values for whole words between 0.85 and 0.95 in the repetitions. We
tested hypothesis-guided planned correlations and predicted that the
word fragment score in one block would correlate with the whole word
score in the next Block. However, these and also almost all of the other
correlations between word fragment scores and subsequent whole words
scores were not significant.
we would like to suggest that it is likely that also the increase of
word fragments consisted of two processes. The first process would be
the rehearsal of the word fragment, and this explains why there were
significant correlations that could increase during the repetition. The
second process would be that increasingly some more new word fragments
were produced, and this relatively free generative process explains why
the correlations were on average lower than for whole words. In the
context of an immediate serial recall task, Couture et al. (2008)
found that repeated learning of visual letter sequences yielded 2,376
response mouse clicks. Of these clicks 938 responses were errors, with
468 repeated errors and 470 new errors. 159 repeated errors were from
the previous block, but 309 errors were from an even earlier block in
the experiment. This indicates that wrong letter sequences were well
remembered in visual LTM beyond the immediate recall context. When
increasing error learning during repetition is not analyzed this could
be mistaken for an absence of correct response learning, while in fact
both correct and wrong responses increase simultaneously (Lafond et al., 2010). Also McClelland (2001)
warned that Hebbian learning may actually strengthen inappropriate
activations if for instance an over-inclusive prototype was generated
Sparse Written Word Representations
partially correct words be similar to Mojibake? A Mojibake of
unintelligible characters emerges when different writing systems clash,
such as Japanese Kanji JIS and the Western Alphabetic code ASCII (Wlodarczyk, 2005). It is even suggested to make PDF word documents safer by using Mojibake (Bakhtiyari et al., 2014).
PDF documents have an upper layer with an image of the text and a lower
layer with the letters that make up the words. It is suggested that a
way to increase PDF security would be to eliminate the letter sequences
and instead of well-sequenced letters only Mojibake would be offered in
the lower layer which would render copying of the PDF document
layers of word memory in participants, and not just in PDF documents.
The current experiment showed that there may be a lower sensory layer
consisting of free sparse word fragments which can be image-like
pictures or acoustic-like sounds and an upper layer that consists of
language-specific, orthographically and semantically constrained words.
This is just the opposite of what was suggested by Chomsky for spoken
language (Chomsky, 1959, 2002; Chomsky and DiNozzi, 1972).
He suggested that we are creative rather than conditioned insofar as
there is a lower layer of deep meaning anchored in action schemata,
while the human mind finds myriads of ways to express the meaning in
syntactic structures on a surface level. However, the current study
shows that when top–down word representations from a mental lexicon
cannot trigger an unfamiliar word from the LTM store because of complete
novelty, or a small constrained lexicon, incomplete sparse sublexical
bottom-up sensory impressions of word input take over (see also Nuerk et al., 2000) which are reinforced over repetitions even if partly wrong.
who wrote an immensely instructive early review about learning with
repetition, in particular the Hebb effect. The Hebb effect shows that
dispersed repeated sequences of letters, digits or words are better
learned than novel sequences in an immediate recall task, even if
participants do not notice the repetition (see also McKelvie, 1987). Frick suggested a recorder model
with a fixed amount of recording tape. Thus, in general, reproduction
of words would show high fidelity of the original word. However, when
too many items are presented, only a small amount of representational
medium could be devoted to each item resulting in a low fidelity
representation. He described that while participants represent a set of
words, they do not represent psychophysical parameters such as duration,
or mimic the pitch, accent, rhythm or loudness (Frick, 1988, p. 223, but see Lange-Küttner et al., 2013).
Instead, an unparsed, uncategorized, more or less degraded input would
need to be recovered for recall. According to Frick, the recovery for
recall would represent a second level of processing which can be
facilitated with grouping or chunking (see also Cumming et al., 2003) into categories or perceptual boundaries of Gestalt-like stimuli and stimulus sequences.
would suggest that in the case of written words, this process of
recovery is not creative but on the contrary, it is conventional insofar
as it is governed via the lexicon that prescribes an exact replication
of the graphic orthography. In terms of working memory, the inner scribe
and the visual cache components of the visual-spatial sketchpad of the
working memory model may be likely candidates for the visual rehearsal
of words fragments. Logie (2011,
p. 214) describes visual rehearsal as follows: ‘The Inner Scribe
component (…) can allow visual codes to be held for longer by mentally
rehearsing the codes held in the Visual Cache.’ Thus, we would suggest
that rehearsal of written word fragments is most likely to take place in
the inner scribe and the visual cache, firstly because participants
held some sparse details of recently perceived unfamiliar words (in the
visual cache), and secondly, during the repetition these were processed
further (in the inner scribe). However, in order to avoid learning wrong
words, an active mapping process would need to take place where the
visual slave systems are controlled by the central executive whether the
visual orthographic code matches LTM representations in the episodic
memory system that stores accumulated conventional orthographic patterns
encountered during previous experiences.
showed that in spoken language, children were drawing an advantage from
more easily pronounceable words in word recognition only from age 10.
This late onset suggests that in word memory children develop
language-specific acoustic and probably also graphonomic sensitivities
relatively late after being taught to read. It also suggests that
increased sensitivities may need an increased categorical filter or
quality control. For instance, children seem to be biased toward
positive feedback whether it is justified or not (Crone et al., 2004; Eppinger et al., 2009; Lange-Küttner et al., 2012)
which may help to persevere in a learning task, but not to discriminate
when words do not ‘look right.’ Moreover, the current study showed that
this is still the case in young adults if they encounter unfamiliar
words with no ready-made word template available for word recall.
Future Research Questions
development, the onset of written language changes word memory because
the new visual modality is added to language. For instance, in beginning
readers, their small lexicon of written words makes them rely heavily
on familiar items in their visual word memory, while the saturated
lexicon of spoken words accumulated over several years allows them to
better memorize novel words (Lange-Küttner and Martin, 1999; Lange-Küttner, 2005; Lange-Küttner and Krappmann, 2011).
Also children with reading difficulties produce significantly more
misspellings that are close visual matches to the target word rather
than phonological mismatches (Lennox and Siegel, 1996).
This is why the current study put more weight on orthographic patterns
in visual word memory than on phonemic sound transitions in spoken word
memory. Visual word rehearsal may be counter-intuitive, but for written
language it is quite a crucial research question that needs further
testing. For instance, while it is a reasonable assumption that word
fragments develop into whole words, the current study did not find any
statistical evidence for a trade-off between word fragments and whole
fragments in free recall is rather worrying. It has indeed been claimed
recently that error learning during repetition may be responsible for
developmental dyslexia (Szmalec et al., 2011).
While learning with repetition was completely absent in dyslexic
participants when they had to remember the places of dots, it was only
attenuated in visual and auditory learning of letter sequences.
Likewise, also children with Down syndrome showed learning with
repetition comparable to normally developing children which explained
their good vocabulary despite a verbal short-term memory deficit (Mosse and Jarrold, 2010).
the current study could not show that word fragments would develop into
a whole words during repeated rehearsals, there is a hint in the
non-significant correlations, which developed from a negative into a
positive correlation (monolingual males and females), or from a positive
into a negative correlation (bilingual males and females) during the
experiment, see Table Table44.
While this appeared to be a smooth trend, none of these correlations
ever reached significance. We also tried to increase the correlations by
distinguishing between word fragments in response to familiar vs.
unfamiliar words, but again without obtaining significant correlations
with whole word responses of the same kind.
comparison with previous research showed some indicators that rehearsal
of whole words and word fragments is based on two different cognitive
processes. Future research could use an item-based methodology where the
fate of an individual word fragment is followed up. For instance, Carey (2010)
assumes that extended mapping with context information produces more
constrained meaning in words that were acquired via fast word mapping.
Hence, extended mapping could facilitate the transition of a response
from the lower free sensory layer to the upper semantically and
orthographically constrained layer. This transformation from a word
fragment to a proper word recall could be tested using the category size
effect (Hunt and Seta, 1984).
This effect demonstrates that words from small categories are better
recalled following orientating relational processing, and words from
large categories are better recalled following individual item
processing. One could envisage an experiment where an increasingly
longer word list in the repetitions gradually provides more context
which could support the refinement of a word fragment into a correct
whole word, or an experiment where a word list gradually becomes more
homogeneous during repetition. For example, if the Word List with EU
towns would gradually change into a Word List with French towns only,
providing a more systematic database, would the first initially
introduced French town that was recalled as a word fragment be spelled
correctly once all town names are presented in the same language? In
this item-based experiment, unbeknown to the participants, only the
rehearsal of the first word fragment would be important, while the
remaining words could be left unscored.
that the current study provided compelling evidence that written word
fragments are likely to be produced when unfamiliar words are
encountered, and that these word fragments are rehearsed and increase
during repetition. We suggest that written word fragments seem to be
free and highly idiosyncratic which currently makes it difficult to
demonstrate how a written word fragment can be rehearsed until a whole
word emerges. We suggest that extended mapping may simultaneously
constrain the semantic content and the orthography of a written word
fragment so that it ‘looks right.’
it is also imaginable that word fragments never develop into proper
words but persist in memory. In the development of young children’s
first spoken word production, invented words were found to be abruptly
dropped in favor of conventional words only (Dromi, 1987).
Anecdotal evidence from children shows that strict rules can control
orthographic output and inhibit the rehearsal activity at the lower
level rather than evolve it. We introduced this study with the neural
network simulation of the Ebbinghaus study (Lange-Küttner, 2011) because Ebbinghaus (1964)
learned the nonsense syllables always to perfection and the gains that
he described were only in terms of time. However, a focus on perfect
accuracy may inevitably simultaneously inhibit the learning potential
with regards to memory for unfamiliar words of any kind. Hence, to
investigate error learning and the interactivity between fragile letter
sequences and robust word representations is an important future
Conflict of Interest Statement
authors declare that the research was conducted in the absence of any
commercial or financial relationships that could be construed as a
potential conflict of interest.
- Baddeley A. D. (1999). Essentials of Human Memory.
Hove: Psychology Press.
- Baddeley A. D. (2012). Working memory: theories, models, and controversies.
Annu. Rev. Psychol.
10.1146/annurev-psych-120710-100422 [PubMed] [Cross Ref]
- Baddeley A. D., Thomson N., Buchanan M. (1975). Word length and the structure of short-term memory.
J. Verbal Learn. Verbal Behav.
10.1016/S0022-5371(75)80045-4 [Cross Ref]
- Bailey T. M., Hahn U. (2001). Determinants of wordlikeness: phonotactics or lexical neighborhoods?
J. Mem. Lang.
10.1006/jmla.2000.2756 [Cross Ref]
- Bakhtiyari K., Salehi H., Embi M. A., Shakiba M., Zavvari A., Shahbazi-Moghadam M., et al. (2014). Ethical and unethical methods of plagiarism prevention in academic writing.
Int. Educ. Stud.
10.5539/ies.v7n7p52 [Cross Ref]
- Beringer J. (1994). Experimental Run Time System (ERTS), Version 3.0 and Experimental Run Time Coding System (ERTSCODE), Version 2.20, Frankfurt.
Available at: www.erts.de.
- Blum D., Yonelinas A. P. (2001). Transfer across modality in perceptual implicit memory.
Psychon. Bull. Rev.
10.3758/BF03196151 [PubMed] [Cross Ref]
- Carey S. (2010). Beyond fast mapping.
Lang. Learn. Dev.
10.1080/15475441.2010.484379 [PMC free article] [PubMed] [Cross Ref]
- Challis B. H., Sidhu R. (1993). Dissociative effect of massed repetition on implicit and explicit measures of memory.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037//0278-73126.96.36.199 [PubMed] [Cross Ref]
- Chomsky N. (1959). A review of BF Skinner’s Verbal Behavior.
10.2307/411334 [Cross Ref]
- Chomsky N. (2002). Syntactic Structures.
Berlin: De Gruyter;
10.1515/9783110218329 [Cross Ref]
- Chomsky N., DiNozzi R. (1972). Language and Mind.
New York, NY: Harcourt Brace Jovanovich.
- Cleary A. M. (2002). Recognition with and without identification: dissociative effects of meaningful encoding.
10.3758/BF03196431 [PubMed] [Cross Ref]
- Cleary A. M., Greene R. L. (2000). Recognition without identification.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037//0278-73188.8.131.523 [PubMed] [Cross Ref]
- Cleary A. M., Greene R. L. (2001). Memory for unidentified items: evidence for the use of letter information in familiarity processes.
10.3758/BF03196405 [PubMed] [Cross Ref]
- Coleman J., Pierrehumbert J. (1997). Stochastic phonological grammars and acceptability.
Available at: http://arxiv.org/abs/cmp-lg/9707017
- Coltheart M. (2012). “Dual-route theories of reading aloud,” in
Visual Word Recognition Volume 1 Models and Methods, Orthography and Phonology, ed.Adelman J. S., editor. (London: Psychology Press; ).
- Connelly V., Dockrell J. E., Walter K., Critten S. (2012). Predicting
the quality of composition and written language bursts from oral
language, spelling, and handwriting skills in children with and without
specific language impairment.
10.1177/0741088312451109 [Cross Ref]
- Couture M., Lafond D., Tremblay S. (2008). Learning correct responses and errors in the Hebb repetition effect: two faces of the same coin.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037/0278-73184.108.40.2064 [PubMed] [Cross Ref]
- Couture M., Tremblay S. (2006). Exploring the characteristics of the visuospatial Hebb repetition effect.
10.3758/BF03195933 [PubMed] [Cross Ref]
- Cowan N. (1995). Attention and Memory: An Integrated Framework.
New York, NY: Oxford University Press.
- Crone E. A., Jennigs J. R., Van der Molen M. W. (2004). Developmental change in feedback processing as reflected by phasic heart rate changes.
10.1037/0012-16220.127.116.118 [PubMed] [Cross Ref]
- Cumming N., Page M., Norris D. (2003). Testing a positional model of the Hebb effect.
10.1080/741938175 [PubMed] [Cross Ref]
- Darling S., Allen R. J., Havelka J., Campbell A., Rattray E. (2012). Visuospatial
bootstrapping: long-term memory representations are necessary for
implicit binding of verbal and visuospatial working memory.
Psychon. Bull. Rev.
10.3758/s13423-011-0197-3 [PubMed] [Cross Ref]
- Darling S., Havelka J. (2010). Visuospatial bootstrapping: evidence for binding of verbal and spatial information in working memory.
Q. J. Exp. Psychol.
10.1080/17470210903348605 [PubMed] [Cross Ref]
- Dehaene S., Pegado F., Braga L. W., Ventura P., Nunes Filho G., Jobert A., et al. (2010). How learning to read changes the cortical networks for vision and language.
10.1126/science.1194140 [PubMed] [Cross Ref]
- Dromi E. (1987). Early Lexical Development.
Cambridge, MA: Cambridge University Press.
- Ebbinghaus H. (1964). Memory. A Contribution to Experimental Psychology, eds
Ruger H. A., Bussenius C. E., Hilgard E. R., editors. (New York, NY: Dover Publications; ).
- Edwards J., Beckman M. E., Munson B. (2004). The
interaction between vocabulary size and phonotactic probability effects
on children’s production accuracy and fluency in nonword repetition.
J. Speech Lang. Hear Res.
10.1044/1092-4388(2004/034) [PubMed] [Cross Ref]
- Eppinger B., Mock B., Kray J. (2009). Developmental differences in learning and error processing: evidence from ERPs.
10.1111/j.1469-8986.2009.00838.x [PubMed] [Cross Ref]
- Fecteau J. H., Munoz D. P. (2003). Exploring the consequences of the previous trial.
Nat. Rev. Neurosci.
10.1038/nrn1114 [PubMed] [Cross Ref]
- Frick R. W. (1988). Issues of representation and limited capacity in the auditory short-term store.
Br. J. Psychol.
10.1111/j.2044-8295.1988.tb02284.x [PubMed] [Cross Ref]
- Friederici A. D., Lachmann T. (2002). “From language to reading and reading disability,” in
Basic Functions of Language, Reading and Reading Disability, eds
Witruk E., Friederici A. D., Lachmann T., editors. (Boston: Kluwer; ), 9–21
10.1007/978-1-4615-1011-6_2 [Cross Ref]
- Gathercole S. E. (2006a). Complexities and constraints in nonword repetition and word learning.
10.1017/S014271640606053X [Cross Ref]
- Gathercole S. E. (2006b). Nonword repetition and word learning: the nature of the relationship.
10.1017/S0142716406060383 [Cross Ref]
- Gathercole S. E., Baddeley A. D. (1993). Working Memory and Language.
- Gathercole S. E., Baddeley A. D. (1997). Sense and sensitivity in phonological memory and vocabulary development: a reply to Bowey (1996).
J. Exp. Child Psychol.
10.1006/jecp.1997.2391 [PubMed] [Cross Ref]
- Gupta P., Tisdale J. (2009). Word learning, phonological short-term memory, phonotactic probability and long-term memory: towards an integrated framework.
Philos. Trans. R. Soc. B Biol. Sci.
10.1098/rstb.2009.0132 [PMC free article] [PubMed] [Cross Ref]
- Horst J. S. (2013). Context and repetition in word learning.
10.3389/fpsyg.2013.00149 [PMC free article] [PubMed] [Cross Ref]
- Horst J. S., Parsons K. L., Bryan N. M. (2011). Get the story straight: contextual repetition promotes word learning from storybooks.
10.3389/fpsyg.2011.00017 [PMC free article] [PubMed] [Cross Ref]
- Horst J. S., Samuelson L. K. (2008). Fast mapping but poor retention by 24-month-old infants.
10.1080/15250000701795598 [Cross Ref]
- Huestegge L., Heim S., Zettelmeyer E., Lange-Küttner C. (2012). Gender-specific contribution of a visual cognition network to reading abilities.
Br. J. Psychol.
10.1111/j.2044-8295.2011.02050.x [PubMed] [Cross Ref]
- Hunt R. R., Seta C. E. (1984). Category size effects in recall: the roles of relational and individual item information.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037/0278-7318.104.22.1684 [Cross Ref]
- International Phonetic Association (2005). The International Phonetic Alphabet (Revised Version).
Available at: https://www.internationalphoneticassociation.org/content/ipa-chart
- Jarrold C., Tam H. (2011). “Rehearsal and the development of working memory,” in
Cognitive Development and Working Memory: A Dialogue Between Neo-Piagetian Theories and Cognitive Approaches,
Barrouillet P., Gaillard V., editors. (Hove: Psychology Press; ), 177–200.
- Kucera H., Francis N. (1967). Computational Analysis of Present-Day American English.
Providence, RI: Brown University Press.
- Lafond D., Tremblay S., Parmentier F. (2010). The ubiquitous nature of the Hebb Repetition Effect: error learning mistaken for the absence of sequence learning.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037/a0018469 [PubMed] [Cross Ref]
- Lange-Küttner C. (2005). Word structure effects in German and British reading beginners.
Z. Pädagog. Psychol.
10.1024/1010-0622.214.171.124 [Cross Ref]
- Lange-Küttner C. (2011). Ebbinghaus simulated: just do it 200 times.
10.1109/DEVLRN.2011.6037363 [Cross Ref]
- Lange-Küttner C. (2014). Do drawing stages really exist? Children’s early mapping of perspective.
Psychol. Aesthet. Creat. Arts
10.1037/a0036199 [Cross Ref]
- Lange-Küttner C., Averbeck B. B., Hirsch S. V., Wießner I., Lamba N. (2012). Sequence learning under uncertainty in children: self-reflection vs. self-assertion.
10.3389/fpsyg.2012.00127 [PMC free article] [PubMed] [Cross Ref]
- Lange-Küttner C., Krappmann L. (2011). Ein
modalitätsspezifisches Wortgedächtnis ist ausreichend:
wortgedächtnismodalität bei Leseanfängern und Neuronalen Netzen (One
modality-specific word memory is sufficient – word memory modality in
reading beginners and neural networks, engl. abstract).
Z. Pädagog. Psychol.
10.1024/1010-0652/a000027 [Cross Ref]
- Lange-Küttner C., Martin S. (1999). Reversed effects of familiarity and novelty in visual and auditory working memory for words.
- Lange-Küttner C., Puiu A.-A., Nylund M., Cardona S., Garnes S. (2013). Speech preparation and articulation time in bilinguals and men.
Int. J. Speech Lang. Pathol. Audiol.
10.12970/2311-1917.2013.01.01.5 [Cross Ref]
- Leech G. N., Rayson P., Wilson A. (2001). Word Frequencies in Written and Spoken English.
Harlow: Pearson Longman.
- Lennox C., Siegel L. S. (1996). The development of phonological rules and visual strategies in average and poor spellers.
J. Exp. Child Psychol.
10.1006/jecp.1996.0022 [PubMed] [Cross Ref]
- Logan G. D. (1990). Repetition priming and automaticity: common underlying mechanisms?
10.1016/0010-0285(90)90002-L [Cross Ref]
- Logie R. H. (2011). The functional organization and capacity limits of working memory.
Curr. Dir. Psychol. Sci.
10.1177/0963721411415340 [Cross Ref]
- McClelland J. L. (2001). “Failures to learn and their remediation: a Hebbian account,” in
Mechanisms of Cognitive Development: Behavioral and Neural Perspectives, eds
McClelland J. L., Siegler R. (Mahwah, NJ: Psychology Press; ), 97–121.
- McKelvie S. J. (1987). Learning and awareness in the Hebb Digits Task.
J. Gen. Psychol.
10.1080/00221309.1987.9711057 [Cross Ref]
- McLeod P., Plunkett K., Rolls E. T. (1998). Introduction to Connectionist Modelling of Cognitive Processes.
Oxford: Oxford University Press.
- Mitchell J. F., Zipser D. (2003). Sequential memory-guided saccades and target selection: a neural model of the frontal eye fields.
10.1016/S0042-6989(03)00468-1 [PubMed] [Cross Ref]
- Mohamed W., Elbert T., Landerl K. (2011). The development of reading and spelling abilities in the first 3 years of learning Arabic.
10.1007/s11145-010-9249-8 [Cross Ref]
- Morra S., Camba R. (2009). Vocabulary learning in primary school children: working memory and long-term memory components.
J. Exp. Child Psychol.
10.1016/j.jecp.2009.03.007 [PubMed] [Cross Ref]
- Mortensen U., Nachtigall C. (2000). Visual channels, Hebbian assemblies and the effect of Hebb’s rule.
10.1007/s004220050594 [PubMed] [Cross Ref]
- Mosse E. K., Jarrold C. (2010). Searching
for the Hebb effect in Down syndrome: evidence for a dissociation
between verbal short-term memory and domain-general learning of serial
J. Intellect. Disabil. Res.
10.1111/j.1365-2788.2010.01257.x [PubMed] [Cross Ref]
- Munson B., Kurtz B. A., Windsor J. (2005). The
influence of vocabulary size, phonotactic probability, and wordlikeness
on nonword repetitions of children with and without specific language
J. Speech Lang. Hear Res.
10.1044/1092-4388(2005/072) [PubMed] [Cross Ref]
- Nuerk H.-C., Rey A., Graf R., Jacobs A. M. (2000). Phonographic sublexical units in visual word recognition.
Curr. Psychol. Lett. Behav. Brain Cogn.
- Nyberg L., Nilsson L.-G., Olofsson U. (1994). Repetition effects on word fragment completion: the role of competition among responses.
Scand. J. Psychol.
10.1111/j.1467-9450.1994.tb00933.x [Cross Ref]
- Oberauer K., Meyer N. (2009). The contributions of encoding, retention, and recall to the Hebb effect.
10.1080/09658210903107861 [PubMed] [Cross Ref]
- Page M. P. A., Cumming N., Norris D., Hitch G. J., McNeil A. M. (2006). Repetition learning in the immediate serial recall of visual and auditory materials.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037/0278-73126.96.36.1996 [PubMed] [Cross Ref]
- Page M. P. A., Cumming N., Norris D., McNeil A. M., Hitch G. J. (2013). Repetition-spacing and item-overlap effects in the Hebb repetition task.
J. Mem. Lang.
10.1016/j.jml.2013.07.001 [Cross Ref]
- Page M. P. A., Norris D. (2009). “Is
there a common mechanism underlying word-form learning and the Hebb
repetition effect? Experimental data and a modelling framework,” in
Interactions Between Short-Term and Long-Term Memory in the Verbal Domain, eds
Thorn A., Page M., editors. (Hove: Psychology Press; ), 136–156.
- Paulesu E., McCrory E., Fazio F., Menoncello L., Brunswick N., Cappa S. F., et al. (2000). A cultural effect on brain function.
10.1038/71163 [PubMed] [Cross Ref]
- Plaut D., McClelland J., Seidenberg M., Patterson K. (1996). Understanding normal and impaired reading: computational principles in quasi-regular domains.
10.1037/0033-295X.103.1.56 [PubMed] [Cross Ref]
- Plunkett K., Marchman V. (1990). From Rote Learning to System Building: Technical Report, Center for Research in Language.
San Diego: University of California.
- Rack J., Hulme C., Snowling M., Wightman J. (1994). The role of phonology in young children learning to read words: the direct-mapping hypothesis.
J. Exp. Child Psychol.
10.1006/jecp.1994.1003 [Cross Ref]
- Samuels S. J., Miller N. L., Eisenberg P. (1979). Practice effects on the unit of word recognition.
J. Educ. Psychol.
10.1037/0022-06188.8.131.524 [Cross Ref]
- Storkel H. L., Armbruster J., Hogan T. P. (2006). Differentiating phonotactic probability and neighborhood density in adult word learning.
J. Speech Lang. Hear Res.
10.1044/1092-4388(2006/085) [PMC free article] [PubMed] [Cross Ref]
- Storkel H. L., Rogers M. A. (2000). The effect of probabilistic phonotactics on lexical acquisition.
Clin. Linguist. Phon.
10.1080/026992000415859 [Cross Ref]
- Szmalec A., Loncke M., Page M. P. A., Duyck W. (2011). Order or disorder? Impaired Hebb Learning in Dyslexia.
J. Exp. Psychol. Learn. Mem. Cogn.
10.1037/e512592013-202 [PubMed] [Cross Ref]
- Treiman R. (1993). Learning to Spell.
Oxford: Oxford University Press.
- Treiman R., Goswami U., Bruck M. (1990). Not all nonwords are alike: implications for reading development and theory.
10.1037/e512592013-202 [PubMed] [Cross Ref]
- Underwood B. J. (1961). Ten years of massed practice on distributed practice.
10.1037/h0047516 [Cross Ref]
- Vitevitch M. S., Luce P. A. (2004). A web-based interface to calculate phonotactic probability for words and nonwords in English.
Behav. Res. Methods Instrum. Comput.
10.3758/BF03195594 [PMC free article] [PubMed] [Cross Ref]
- Whitney P. (1998). The Psychology of Language.
Boston: Houghton Miﬄin.
- Wlodarczyk A. (2005). Japanese – An International Language. Paper Presented at the Paris Lectures in Japanese Linguistics.
Lille: Université Charles de Gaulle.
- Ziegler J. C., Goswami U. (2005). Reading acquisition, developmental dyslexia, and skilled reading across languages: a psycholinguistic grain size theory.
10.1037/0033-2909.131.1.3 [PubMed] [Cross Ref]
Mojibake – The rehearsal of word fragments in verbal recall