Language Learning and Development
ISSN: 1547-5441 (Print) 1547-3341 (Online) Journal homepage: http://www.tandfonline.com/loi/hlld20
Twenty-Four-Month-Olds’ Perception of WordMedial Onsets and Codas
Yuanyuan Wang & Amanda Seidl
To cite this article: Yuanyuan Wang & Amanda Seidl (2016) Twenty-Four-Month-Olds’
Perception of Word-Medial Onsets and Codas, Language Learning and Development, 12:4,
447-460, DOI: 10.1080/15475441.2016.1150185
To link to this article: http://dx.doi.org/10.1080/15475441.2016.1150185
Published online: 04 Apr 2016.
Submit your article to this journal
Article views: 54
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=hlld20
Download by: [Nationwide Childrens Hospital]
Date: 20 October 2016, At: 10:44
LANGUAGE LEARNING AND DEVELOPMENT
2016, VOL. 12, NO. 4, 447–460
http://dx.doi.org/10.1080/15475441.2016.1150185
Twenty-Four-Month-Olds’ Perception of Word-Medial Onsets and
Codas
Yuanyuan Wanga and Amanda Seidlb
a
Linguistics, Purdue University; bSpeech, Language and Hearing Sciences, Purdue University
ABSTRACT
Recent work has shown that children have detailed phonological representations of consonants at both word-initial and word-final edges.
Nonetheless, it remains unclear whether onsets and codas are equally
represented by young learners since word edges are isomorphic with
syllable edges in this work. The current study sought to explore toddler’s
sensitivity to onsets and codas without this overlap. This question is of
theoretical interest since many theories of phonology suggest that onsets
may be weighted more heavily than codas. We examined English-learning
24-month-olds’ sensitivity to mispronunciations in word-medial onset and
coda positions in a word learning task. We predicted that if children’s
sensitivity to onsets and codas is positionally constrained, then they should
be more tolerant of Coda mispronunciations than onset ones. The results
supported this prediction: children were less sensitive to coda mispronunciations than to onset mispronunciations in word-medial position, suggesting positionally driven biases in word processing.
Introduction
A large literature documents adults’ rapid and robust sensitivity to phonological distinctions in word
recognition (Magnuson, Dixon, Tanenhaus, & Aslin, 2007; McMurray, Tanenhaus, Aslin, & Spivey,
2003). This sensitivity enables adults to focus on the linguistic dimensions that are critical for
distinguishing among highly similar words. For example, this enables adults to discriminate margin
from Marvin by attending to a change in the word-medial onset and to discriminate grateful from
graceful by attending to a change in the word-medial coda. Despite their well-specified representations of words, adults are also able to accommodate variability in the words produced by different
talkers or at different speech rates. For example, after less than a minute of familiarization to an
accent which significantly impacts the pronunciation of phonemes, English-speaking adults are able
to recognize accented English words produced by Spanish and Chinese speakers (Clarke & Garrett,
2004). Similarly, prior studies show that children are also able to accommodate small acoustic
differences due to accent in certain situations (e.g., Schmale, Cristia, & Seidl, 2012). However,
these abilities are limited and may rely heavily on lexical knowledge (White & Morgan, 2008), and
the exact nature and specificity of infant lexical representations is far from clear. We would ideally
like to know if children bring certain phonological biases in encoding words such that while they
accommodate some types of mispronunciations, they will not accommodate others.
The primary purpose of this study is to investigate one such area of phonological structure,
namely, syllable structure. Specifically, we examine whether young children show an asymmetrical
pattern in processing word-medial onsets and codas while learning novel words. This is an
important question as many phonological theories, motivated by surveys of language typology,
CONTACT Yuanyuan Wang
[email protected]
Linguistics Program, Purdue University, West Lafayette, IN 47906.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/hlld.
© 2016 Taylor & Francis
448
Y. WANG AND A. SEIDL
suggest such biases exist because, cross-linguistically, languages favor onsets over codas (e.g., Blevins,
1996; Prince & Smolensky, 1993). In support of this idea, evidence from children’s production shows
a clear developmental pattern regarding the acquisition of syllable types and positions, suggesting
that onsets are acquired earlier than codas (e.g., Demuth, 1995; Levelt, 2012; C. C. Levelt, Schiller, &
Levelt, 2000). In addition to the evidence from production, the speech perception literature has
demonstrated that pre-literate children show different degrees of phonological awareness of onsets
and rhymes (Goswami & Bryant, 1990); this evidence, to a certain degree, can be related to children’s
different level of sensitivity to onsets and codas. Even as mature language learners, adults may still
find codas to be more difficult to identify than onsets. The asymmetry between onsets and codas may
be related to phonologically driven learning biases. Thus, investigation of children’s sensitivity to
word-medial onset and coda mispronunciations will help to illuminate the nature of children’s early
lexical representations and the presence or absence of positional biases.
Much like adults, researchers have found that infants are able to distinguish small phonetic
contrasts from the second half of their first year. For example, Jusczyk and Aslin (1995) showed that
even 7.5-month-old infants recognized the difference between cup and tup, indicating that these
young infants have a fairly well-specified representation of word-initial onset segments. Subsequent
work exploring the level of detail in young children’s phonological representations has similarly
found that they are able to distinguish between canonical word forms and mispronounced forms.
This sensitivity to mispronunciations has been found for word-initial onset consonants at 12 months
(Ballem & Plunkett, 2005; Fennell & Werker, 2003; Mani & Plunkett, 2010; Nazzi, 2005; Swingley,
2005, 2009; White & Morgan, 2008), vowels at 12 months (Mani & Plunkett, 2007, 2010), wordmedial onset consonants at 19 months (Swingley, 2003), and word-final coda consonants at
17 months (Levelt, 2012; Swingley, 2005, 2009) for both familiar and novel words.
Nevertheless, up until now, only a few studies have looked at both word-initial onset and wordfinal coda and these have yielded mixed results. Some studies have shown that children are sensitive
to segmental contrasts regardless of the word position (e.g., Nazzi, 2005; Swingley, 2009; Zamuner,
2013). For example, 14–22-month-old (mean age: 17;23) English-speaking children fixated named
targets more when hearing correct pronunciations than hearing mispronunciations either when the
mispronunciations involved word-initial or word-final position (Swingley, 2009). In a categorization
task, 20-month-old French-learning children showed similar sensitivity to phonological distinctions
in both word-initial and word-final positions (Nazzi, 2005). In addition, 29-month-old Dutchlearning children in Zamuner (2013)’s study did not show differential perception of segmental
contrasts in word-initial and -final positions. In contrast, others studies have found that children
are better at identifying differences in word-initial position than in word-final positions (e.g.,
Altvater-Mackensen & Fikkert, 2010). For example, Altvater-Mackensen and Fikkert (2010) found
that 14-month-old Dutch learners showed earlier sensitivity to the stop-fricative contrast in wordinitial compared with word-final position in newly learned words. These conflicting results may due
to the age differences in the children being tested. Children who showed recognition of differences in
both word-initial and word-final positions in the above-mentioned studies were much older and
may be at a more mature stage of language development. Indeed, the role of age or developmental
stage on the perception of word-initial and word-final consonants has been found in numerous
studies. For example, Levelt (2012)’s study demonstrated that 18-, but not 14-month-olds, are
sensitive to word-final consonant omission in a novel-word learning task. In the same vein,
Zamuner (2006) found negative results for word-final mispronunciations with 11-month-olds, but
positive results for some phonological contrasts in this position with 16-month-olds. In summary,
these studies revealed that up until 14 months of age, children showed a more developed ability to
process contrasts in word-initial positions than in word-final positions.
However, it is noteworthy that the majority of the previous studies mentioned have only
focused on mispronunciations in monosyllabic words, in which word edges are isomorphic with
syllable edges. It is, however, important to separate out word-edge effects from syllable position
effects, which has not been done in any of these studies cited above. This is crucial because we
LANGUAGE LEARNING AND DEVELOPMENT
449
know that word-edges are both psychologically (because of primacy and recency effects (Capitani,
Della Sala, Logie, & Spinnler, 1992)) and phonologically special. For example, 7-month-olds
identify words by focusing on the first and last syllables, but ignoring medial syllables
(Benavides-Varela & Mehler, 2015). Similarly, adult show better learning of phonotactic regularities at word-edge than at word-medial positions. Specifically, they learned that C1 and C2 came
from distinct sets in structures of C1VCCVC2, where the target consonants were at word-edge
positions, but failed to learn this constraint in words of CVC1C2VC structure, where the target
consonants were in word-medial positions (Endress & Mehler, 2010).
In addition, children’s processing of word-initial and word-final positions is not necessarily a
reflection of their ability in encoding onsets and codas, but could be rather a reflection of temporal
processing. This is because, in monosyllabic words, onsets always come before codas. Thus, children’s better knowledge of onsets over codas (as suggested in the work previously mentioned from
production) can also be interpreted as support for the fact that their interpretation of linguistic units
is incremental. However, in word-medial positions, we can avoid such large temporal differences and
even place coda constraints before onset ones. For example, in the monosyllabic word cat /kæt/, the
word-initial sound /k/ occurs before the word-final sound /t/; however, in the bisyllabic word faithful
/feθ.fʊl/, the word-medial onset /f/ occurs after the word-medial coda /θ/. Thus, if children’s
representation of segments is incremental, then they should be more sensitive to word-medial
codas than word-medial onsets. On the contrary, if children’s knowledge of onsets and codas were
influenced by the phonological knowledge rather than a reflection of temporal effect, then we would
expect them to be more sensitive to word-medial onsets than codas.
Finally, it should also be noted that acoustic saliency differs as a function of word position. More
specifically, in monosyllabic words, word-final consonants are acoustically less salient than wordinitial consonants. For example, in English, word-final stops are produced with an unreleased burst
(Denes, 1955) while word-initial ones are produced with aspiration. Thus, the asymmetrical perceptual pattern in onset and coda position seen in the literature may be caused by acoustic differences.
However, the acoustic advantage of onsets relative to codas may be largely reduced in word-medial
position, especially in bisyllabic trochees where the word-medial coda is in the stressed syllable, while
the word-medial onset is located in the unstressed syllable.
Thus, for all the above reasons, it is necessary to look at word-medial position in order to explore
the question of whether there are phonologically driven positional constraints on word representations. Up until now, only a few studies have examined children’s sensitivity to word-medial
consonants at all. For example, Swingley (2003) found that 19-month-old Dutch-learning children
showed better recognition of correct pronunciations over mispronunciations when the word-medial
consonant was altered in familiar words (where for required lesser attention because it was familiar),
however, he only tested children on familiar words with a CVCV structure (with the underscored
consonant as the target mispronunciation). Thus, infants were only tested on mispronunciations of
word-medial onsets, but not on word-medial codas. Using the Headturn Preference procedure (HPP;
Jusczyk & Aslin, 1995), Wang and Seidl (2015) compared infants’ sensitivity to both onsets and
codas in word-medial positions. They found that 12-month-olds learned the experimental pattern
when the target fricatives were licit in the word-medial onset position, but not in the word-medial
coda position. However, by 15 months of age, infants were also able to learn and generalize such
patterns for word-medial codas. Similarly, in a recent study, Archer, Zamuner, Engel, Fais, and
Curtin (2016) explored English-learning 12- and 20-month-olds’ discrimination of coda consonants
in word-medial positions using a habituation paradigm. The results showed that 20-month-olds, but
not 12-month-olds, discriminated this contrast for both voiced and voiceless stops. Nevertheless,
these studies do not tell us much about how well the information in the onsets and codas are
represented while learning novel words, because infants and children often demonstrate less sensitivity to phonological details in words, particularly when they need to simultaneously pay attention
to meaning and form (Werker, Fennell, Corcoran, & Stager, 2002).
450
Y. WANG AND A. SEIDL
The secondary goal of the research was to investigate the degree of phonological deviation in
children’s ability to process segmental contrasts. Many studies have demonstrated that children are
sensitive to the major dimensions of phonological variation, such as voicing, place, and manner,
which define phonological features (e.g., Mani & Plunkett, 2010; Miller & Eimas, 1983; White &
Morgan, 2008). This knowledge is of particular importance to infants and young children who are
faced with the challenging task of building the lexicon of their native language, in which words with
minimal phonetic differences may be abundant. For example, as mature language learners, they will
have to distinguish between cat and pat, which differ only in the place of articulation (POA) of the
first sound. Previous research has found that toddlers in the second year of life have demonstrated
sensitivity to even one such featural change in familiar words (Ballem & Plunkett, 2005; Mani &
Plunkett, 2010; Swingley & Aslin, 2000, 2002). Despite this seemingly fine-grained sensitivity to
featural changes, literature concerning the degree of phonological deviation in children’s word
recognition or processing has been inconclusive. Specifically, in some studies, children up to
23 months responded similarly to single-feature and multiple-feature mispronunciations (Bailey &
Plunkett, 2002; Swingley & Aslin, 2002). In other studies this was not the case. For example, in White
and Morgan (2008) 19-month-olds showed graded sensitivity to the varying degree of phonological
mismatch, and this sensitivity can be generalized to a varying combinations of features. These
conflicting results may arise from differences in task demands, age at test, as well as the very nature
of tasks. In order to add to the existing body of literature on this standing issue, we incorporated
another variable, namely 1-feature and 2-feature change from correct pronunciation, in our experimental design.
In summary, the primary goal of the current research was to explore whether there is an
asymmetry in children’s representation of word-medial onsets and codas in a word learning task.
The secondary goal was to examine whether children’s ability to encode contrasts is affected by the
degree of phonological deviation (1-feature vs. 2-feature change). Specifically, using the Preferential
Looking paradigm (Golinkoff, Hirsh-Pasek, Cauley, & Gordon, 1987), we tested 24-month-old
English-learning children’s response to correct pronunciations (CPs) and mispronunciations to
newly learned words in both onset (MP-Onset) and coda (MP-Coda) positions. We varied the
degree of phonological deviation such that half of the children received 1-feature change from target
labels to the MPs, and the other half received 2-feature changes from the target labels to the MPs.
Children’s eye movements in response to stimuli were monitored and analyzed. If children’s word
representations were constrained by phonologically driven positional biases, then they should show
better recognition of the target object when the label was mispronounced in coda over onset
position. With regard to the secondary research question, we predicted that if children are sensitive
to the degree of phonological deviation, they should display more sensitivity to MPs that deviate
from CPs in two phonological features than to MPs that deviate from CPs in one phonological
feature.
Methods
Participants
Thirty-two healthy full-term monolingual English-learning children (12 females) participated in the
study. These children were between the ages of 21.68 and 25.59 months (M = 23.76 months,
SD = 0.79). An additional 11 children were tested whose results were not reported for the following
reasons: 7 for fussing or crying; 3 for experimenter error; 1 for having looking times with difference
scores (Target-Distractor) more than 2.5 standard deviations off the overall mean.
Children were compensated with a book or a toy for their participation. Informed consent was
given to the caregivers before testing was carried out. The caregivers were also asked to complete the
short form A of the MacArthur-Bates Communicative Developmental Inventory: Level II
Vocabulary Checklist (CDI, Fenson et al., 2000) in order to elicit participants’ productive vocabulary
LANGUAGE LEARNING AND DEVELOPMENT
451
sizes. The number of words produced by these 32 children ranged from 14 to 100 (M = 55.41,
SD = 26.24).
Stimuli
The visual stimuli were digitized pictures of novel objects on a white background, presented on a
large screen TV monitor. Pictures were of similar sizes with bright colors. An example stimulus pair
is shown in Figure 1.
The auditory stimuli were novel bisyllabic trochees (CVC.CV) composed of English phonemes.
These auditory stimuli were produced by a young adult female native speaker of American English
in an infant-directed register. The speaker had experience in recording stimuli for previous infant
studies. She was informed that all the pseudowords involved were trochaic words with a CVC.CV
structure.1 The stimuli were recorded using a wireless Lavalier microphone (AKG WMS40) in a
sound-shielded booth with a Marantz Professional Solid State Recorder (PMD660ENG). The auditory stimuli were digitized at a 44.1 kHz sampling rate and normalized at an amplitude of 75 dB.
The target labels presented to infants in the training trials were: /gæz.bi/, /læz.di/, /dul.mi/, /ruv.
ni/, /rin.va/, /din.la/, /lad.vu/, /dim.la/, and /gab.vu/. Each infant was taught two words, one from the
Onset-group, and the other from the Coda-group. The word from the Onset-group was mispronounced in the word-medial onset position (MP-Onset; CVC.CV; the underlined segment was
mispronounced), and the word from the Coda-group was mispronounced in the word-medial
coda position (MP-Coda; CVC.CV, the underlined segment was mispronounced), in other words,
they received MP-Onset and MP-Coda test trials for different words. In addition to the mispronunciations (MPs) for the two labels, children also received correct pronunciations (CPs) of the two
target labels. For example, the CP and MP-Onset of the target word /gæz.bi/ were /gæz.bi/ and /gæz.
mi/, and the CP and MP-Coda of the target word /lad.vu/ were /lad.vu/ and /lag.vu/, respectively.
Figure 2 presents an example of experimental design. Since we also wanted to explore whether this
effect is more salient when the MPs involved more phonological features, we manipulated the stimuli
in this way so that half the participants received MPs whose mispronounced segments deviated from
the correct pronunciations in only one phonological feature (1-feature change), and the other half in
two phonological features (2-feature change). The number of feature change, duration, amplitude,
and maximum pitch of target labels in the CPs, MP-Onset, and MP-Coda Test trials are given in
Table 1 and Table 2. Separate One-way ANOVAs were conducted to compare these acoustic
measures among the three types of labels. The results showed that there was no systematic difference
in the duration, F(2, 22) = 2.45, p = .110; however, we found a significant main effect of Type (CP,
MP-Onset, MP-Coda) in the amplitude, F(2, 22) = 33.08, p < .001, η2 = .182, and the maximal pitch,
F(2, 22) = 46.14, p < .001, η2 = .808. Post hoc pairwise comparisons with Bonferroni correction
(multiplied p-values by 3) were conducted. The results revealed that MP-Onset and MP-Coda labels
were significantly higher in both measures of amplitude, t(14) = 6.18, p < .001, t(14) = 8.50, p < .001,
Figure 1. Sample visual stimulus pair.
452
Y. WANG AND A. SEIDL
Word
Trial
Visual Stimuli
Salience
No audio
Training (3
trials)
“Look, it’s the gæz.bi. The gæz.bi!”
Test:
MP-Onset
“Where is the gæz.bi? Can you find
the gæz.bi?”
“Where is the gæz.mi? Can you
find the gæz.mi?”
“Where is the gæz.bi? Can you find
the gæz.bi?”
“Where is the gæz.mi? Can you
find the gæz.mi?”
Salience
No audio
Training (3
trials)
“Look, it’s the lag.vu. The lag.vu!”
Test: CP
gæz.bi
Test:
MP-Onset
Test: CP
“Where is the lag.vu? Can you find
the lag.vu?”
“Where is the lad.vu? Can you find
the lad.vu?”
“Where is the lag.vu? Can you find
the lag.vu?”
“Where is the lad.vu? Can you find
the lad.vu?”
Test: CP
lag.vu
Auditory stimuli
Test:
MP-Coda
Test: CP
Test:
MP-Coda
Figure 2. Experimental design.
Table 1. Number of feature changes, duration (ms), maximum pitch (Hz), and intensity (dB) of the CPs and MPs for words from the
Onset-group.
CPs
Feature
1
1
1
1
2
2
2
2
Label
gaz.bi
laz.di
dul.mi
ruv.ni
rin.va
din.la
lad.vu
gab.vu
Dur
(ms)
612
676
546
577
530
532
621
610
MPs-Onset
Amp
(dB)
74
72
74
74
70
73
74
73
F0
(Hz)
346
363
335
488
318
331
378
323
Label
gaz.mi
laz.gi
dul.bi
ruv.di
rin.da
din.ba
lad.gu
gab.mu
Dur
(ms)
580
680
584
608
597
583
722
581
Amp
(dB)
75
78
80
79
76
77
77
80
F0
(Hz)
410
595
581
623
543
573
549
630
and maximal pitch, t(14) = 6.77, p < .001, t(14) = 10.48, p < .001, than CP labels. However,
importantly, there was no significant difference between MP-Onset and MP-Coda labels for both
the measures of the amplitude, t(15) = 0.17, p = 1, and maximal pitch, t(15) = 0.99, p = 1. 2
Procedure
Participants were trained and tested using the Preferential Looking Procedure (Golinkoff et al., 1987)
in a single-walled sound booth. The participants sat on a caregiver’s lap and watched images
projected on the display at eye level. Each caregiver wore darkened sunglasses to avoid looking at
the display, and was asked to avoid interfering with their child’s performance. Speech stimuli were
presented through a central speaker beneath the display. A camera below the display recorded
toddlers’ looking patterns.
LANGUAGE LEARNING AND DEVELOPMENT
453
Table 2. Number of feature changes, duration (ms), maximum pitch (Hz), and intensity (dB) of the CPs and MPs for words from the
Coda-group.
CP
MPs-Coda
Feature
Label
Dur
(ms)
Amp
(dB)
F0
(Hz)
Label
Dur
(ms)
Amp
(dB)
F0
(Hz)
1
1
1
1
2
2
2
2
lad.vu
gab.vu
rin.va
dim.la
dul.mi
ruv.mi
gaz.bi
laz.di
621
610
530
534
546
613
612
676
74
73
70
73
74
73
74
72
378
323
318
383
335
451
346
363
lag.vu
gam.vu
rid.va
dibla
dub.mi
rud.mi
gam.bi
lag.di
741
699
662
547
579
585
671
662
78
79
78
78
78
77
76
79
582
578
615
598
529
565
622
630
The toddlers were taught two novel target words. For each target word, they received 8 trials: 1
Salience trial, 3 Training trials, and 4 Test trials (2 CP trials, and 2 MP trials). Thus, each child
received 16 trials (8 trials x 2 words) in total. Salience trials served to reduce the difference in
exposure time to the two objects, one of which was presented repeatedly during the Training (the
“Target”), while the other (the “Distractor”) only appeared at Test. An attention-getter (smiling
baby’s face) appeared preceding each trial to attract toddlers’ attention. When the toddlers looked at
the attention-getter, the experimental trials began.
In Salience trials, the Target and the Distractor were shown on the right and the left sides of the
screen with no auditory stimulus. During each of the three Training trials, participants saw a picture
(Target) shown in the middle of the screen, while listening to the target label, recorded within a
carrier sentence. The sentences presented during Training trials were: “Look, it’s the [Target]. The
[Target]!” During the four Test trials, the Target and the Distractor were shown simultaneously on
the screen, as in Salience trials. The sentences presented in the test trials were, “Where is the [CP]/
[MP]? Can you find the [CP]/[MP]?” The Test trials were always played in the order of CP-MP-CPMP. This is because we expected that the task of representing word-medial segments while learning
new words might be challenging for young children, and to have CPs come before MPs might
reinforce their learning of these words with a complex syllable structure. Since our primary question
of interest was not in comparing CPs to MPs, but rather in exploring potential differences between
the two types of MPs (MP-Onset, MP-Coda), we were not concerned about creating biases with this
strict ordering.
Speech stimuli began 500 ms after the visual stimuli were shown on the display; the speech stimuli
for each trial was 3500 ms in total; the visual stimuli remained on the display for another 1500 ms
after the speech stimuli were completed. Thus, each trial was 5.5 s. A total of 32 stimuli orders were
created by balancing the target objects, sides, and the location of the mispronounciation (Onset or
Coda).
Coding
Videos of the participants’ looking patterns to objects on the display were digitized at 30 frames per
second and coded offline frame by frame by a highly trained coder using SuperCoder software
(Hollich, 2008). Participants’ looking directions to the left, right, or away, were coded by a skilled
coder. This procedure allowed us to determine the total amount of time and longest looking time
children spent looking to the Target and Distractor.
The Test trials were divided into two phases: a pre-naming phase and a post-naming phase. The
time window before the onset of the target words in each Test trial was the pre-naming phase.
Children’s looking patterns during the pre-naming phase provide a baseline of target preference. In
the post-naming phase, following previous research (Swingley & Aslin, 2000), looking time to each
object in each Test trial was measured in a time window from 367–2000 ms from the beginning of
454
Y. WANG AND A. SEIDL
Table 3. Mean (SD) of PTL and PLL measures separated by naming and trial type.
Trial Type
Measure
PTL
PLL
Phase
Pre-naming
Post-naming
Difference
Pre-naming
Post-naming
Difference
CP
0.444 (0.18)
0.590 (0.14)
0.146 (0.25)
0.442 (0.18)
0.606 (0.14)
0.164 (0.25)
MP-Onset
0.508 (0.33)
0.406 (0.24)
−0.102 (0.40)
0.509 (0.33)
0.388 (0.26)
−0.121 (0.39)
MP-Coda
0.444 (0.32)
0.584 (0.26)
0.140 (0.41)
0.444 (0.32)
0.595 (0.28)
0.151 (0.45)
the target segment, i.e., the beginning of the fourth segment of CVC.CV (MP-Onset), and the third
segment of CVC.CV (MP-Coda). This ensured that any change in the looking pattern occurred as
result of the response to the spoken word. For dependent measures, following previous research
(Mani & Plunkett, 2008; Swingley, 2009), we calculated the proportion of time that participants
looked at the named target picture relative to the total looking time (either at the Target or the
Distractor)—Proportion of Target Looking measure (PTL). In addition, we also calculated children’s
longest looking time to the Target and Distractor within the same time window. The Proportion of
Longest Looking measure (PLL) was calculated by dividing the longest time children spent looking at
the Target by the sum of the longest looking to the Target and the Distractor. Given that children
were tested with correct pronunciations in both CP-Onset and CP-coda trials, the average CPs were
computed over the CP-Onset and CP-Coda, thus resulting in three different types of Test trials (CP,
MP-Onset, MP-Coda) in the analyses. Table 3 presents the PTL and PLL measures for these three
types of test trials in both the pre-naming and post-naming phases.
Results
A repeated measures ANOVA on PTL with Feature (1-feature change, 2-feature change) as betweensubjects factor, Naming (pre-naming, post-naming) and Trial type (CP, MP-Onset, MP-Coda) as
within-subjects factors revealed a significant interaction of Naming and Trial type, F(1.72, 51.46,
Huynh-Feldt corrected values) = 4.96, p = .014, η2 = .142. The only factor that approached
significance was Naming, F(1, 51.46, Huynh-Feldt corrected values) = 2.94, p = .097. However, all
other main factors or interactions were not significant, Fs < 2.62, ps > .116. To explore the source of
the interaction, post hoc paired-samples t-tests were conducted to compare children’s target looking
between pre-naming and post-naming phases. Figure 3 presents the effect of Naming using the PTL
measure for the CP, MP-Onset, and MP-Coda Test trials. In CP trials, the results showed a
significant increase in target looking from the pre-naming to the post-naming phase, t(31) = 3.31,
p = .002 (all reported t-tests were two-tailed). This significant target preference provided evidence
that the toddlers learned the novel words used in the experiment. In MP-Coda trials, the PTL was
marginally higher in the post-naming than in the pre-naming phase, t(31) = 1.94, p = .062, however,
in MP-Onset trials, no significant difference was observed between the pre- and post-naming phases,
t(31) = −1.44, p = .161.
In order to corroborate these results using a different, potentially more sensitive measure, we also
conducted a repeated measures ANOVA on the PLL measure with Feature (1-feature change, 2feature change) as between-subjects factor, Naming (pre-naming, post-naming) and Trial type (CP,
MP-Onset, MP-Coda) as within-subjects factors. The results revealed a significant interaction of
Naming and Trial type, F(1.71, 51.26, Huynh-Feldt corrected values) = 5.99, p = .007, η2 = .166. The
main factor of Naming was marginally significant, F(1, 51.26, Huynh-Feldt corrected values) = 3.09,
p = .089. All other main factors or interactions were not significant, Fs < 2.12, ps > .137. Post hoc
paired-samples t-tests demonstrated that in CP trials, the PLL measure was significantly higher in
the post-naming phase than in the pre-naming phase, t(31) = 3.70, p = .001; In MP-Coda trials, the
PTL was marginally higher in the post-naming phase than in the pre-naming phase, t(31) = 1.91,
LANGUAGE LEARNING AND DEVELOPMENT
0.25
0.2
*
455
+
Effect of naming
PTL measure
0.15
0.1
0.05
CP
MP-Onset
0
MP-Coda
–0.05
–0.1
–0.15
–0.2
* The effect of naming was significant, p < .05.
+ The effect of naming was marginally significant, .05 < p < .1.
Figure 3. Effect of naming in CP, MP-Onset, and MP-Coda trials using PTL measure. Error bars indicate standard error.
p = .066. However, no significant difference was observed between the pre- and post-naming phases
in the MP-Onset trials, t(31) = −1.77, p = .086. Figure 4 presents the effect of naming using the PLL
measure.
It should be noted that the effect of naming (the magnitude of change from pre-naming to postnaming) could be influenced by the measures taken at the baseline, such that they determined how
much room there was to move from the pre-naming phase to the post-naming phase. In order to
address this potential confound, repeated measures ANOVAs on PTL and PLL with Feature (1feature change, 2-feature change) as between-subjects factor, and Trial type (CP, MP-Onset, MPCoda) as within-subjects factors were conducted for pre-naming phase. No significant main effects
or interactions were found using either PTL, Fs <.94, p > .382, or PLL, Fs < .93, p > .387.
Taken together, these results suggest that: (1) differences between pre- and post-naming phases
following CP and MP-Coda that we found in the previous set of analyses were indeed attributable to
the effect of naming;3 and (2) children recognized the target objects when hearing correctly
pronounced labels and labels with mispronounced codas, however, they failed to recognize target
objects when hearing labels with mispronounced onsets.
Discussion
The primary goal of the current study was to explore syllable positional effects on children’s
sensitivity to the mispronunciation of recently learned words. Although a considerable body of
research has documented infants’/children’s sensitivity to mispronunciations in both word-initial
and word-final positions, no previous study has systematically examined whether children show
asymmetrical patterns in recognizing mispronunciations in word-medial onset vs. coda positions.
Our study suggests that children show a positional asymmetry while learning novel words that is
unrelated to word edge effects. Specifically, children looked longer at the target object when the label
was correctly pronounced or was mispronounced in word-medial coda, but failed to do so when the
mispronunciation occurred in word-medial onset position. The secondary goal of this research was
456
Y. WANG AND A. SEIDL
0.25
*
+
0.2
Effect of naming
PLL measure
0.15
0.1
0.05
CP
MP-Onset
0
MP-Coda
–0.05
–0.1
–0.15
–0.2
* The effect of naming was significant, p < .05.
+ The effect of naming was marginally significant, .05 < p < .1.
Figure 4. Effect of naming in CP, MP-Onset, and MP-Coda trials using PLL measure. Error bars indicate standard error. Error bars
indicate standard error.
to examine the degree of phonological deviation in children’s ability to process segmental contrasts.
Previous research examining this question has yielded mixed results. Our study shows that children
are equally sensitive to MPs that deviate from CPs in one phonological feature and two phonological
features in lexical processing.
With respect to our primary research question, the results are consistent with prior studies
showing an asymmetrical learning of phonotactic patterns restricted to either word-medial onset
or coda position in younger infants (Wang & Seidl, 2015). While this finding, at first glance, seems to
contrast with Swingley (2009) and Nazzi (2005)’s studies reporting a symmetrical representation of
word-edge onsets and codas in children at younger ages, there is of course a crucial difference: word
edges are special. In addition, given the task of word learning is cognitively more demanding than
the task of familiar word recognition, it is possible that children’s symmetrical sensitivity to wordinitial and -final mispronunciations in Nazzi (2005) and Swingley (2009)’s work is the result of a
ceiling effect. However, while from the second year on children are able to represent detailed
phonological information correctly in both word-initial and word-final positions of familiar
words, when exposed to a more challenging word learning task in which mispronunciations occur
in word-medial positions, children show onset biases. These biases may appear due to the greater
cognitive demands of complex syllable structure in word learning or may simply reflect the lesser
salience of word-medial positions, either way, it is clear that not all positions may be treated equally.
One may argue that children’s better representation of sounds in onset position over coda
position in our study might be due to the fact that onsets are acoustically more salient than codas.
However, this is less likely in word-medial position in English trochees in which codas appear in the
stressed syllable, and onsets in the unstressed syllable. Indeed, in a recent study, Wang and Seidl
(2015) compared the acoustics of word-medial onsets and codas in bisyllabic trochees with the
structure CVC.CVC, and found that the sounds in the word-medial codas were greater in both
amplitude and F0 than the word-medial onsets. Thus, the acoustic advantage of onsets over codas
LANGUAGE LEARNING AND DEVELOPMENT
457
that is typical at word-edges disappears in these word-medial positions, allowing us to disentangle
positional effects with acoustic saliency.
While disentangling the word-edge effects, sequence effects, as well as acoustic saliency of onsets
and codas, this study reveals for the first time children’s asymmetrical sensitivity to onsets and codas
in the word-medial position in newly learned words. One possible interpretation of this processing
advantage of onset over coda position is that children may have built more detailed phonological
representations of sounds appearing in onset position. Another possibility is that children may have
built detailed phonological representations of sounds in both onset and coda positions, however,
they are more willing to accommodate a change in coda position than in onset position, such that
mispronunciations in onset, but not in coda position, hinder word recognition. Interestingly, the
asymmetrical processing of word-medial onsets and codas may account for why the world’s
languages are the way they are. Specifically, despite the range of syllable structures present in the
languages of the world, typological investigations have demonstrated that CV is the only syllable type
that is always licit across the world languages. Given the complexity of language structures to be
learned by children, a potential role of onset biases are to direct learners’ attention to the kind of
information that is needed to further develop their perceptual capacity.
With respect to our secondary research questions, we do not find graded sensitivity to the degree
of phonological deviation on word recognition. This result is in line with Bailey and Plunkett (2002)
and Swingley and Aslin (2002)’s findings that children up to 23 months responded similarly to
single-feature and multiple-feature mispronunciations. However, our results differ from White and
Morgan (2008)’s findings that 19-month-olds showed linearly graded sensitivity to mispronunciations involving one, two and three phonological features. Specifically, children showed significant
increases in looking toward the familiar object in the correct and one-feature conditions, but not in
two- or three-feature conditions. The discrepancies between our studies may be due to the referential
status of the visual competitors. Specifically, they presented children with a familiar and an
unfamiliar object. In contrast, children were presented with novel objects and novel labels in our
study. It is also likely that since the children in our study were older, their phonological representation of newly learned words is more detailed. Finally, another possibility is that the differences in the
feature changes involved in the two studies result in different magnitudes of acoustic-phonetic
mismatch; however, only future studies with systematic investigations can tease these explanations
apart.
Last but not least, this study adds to the existing body of literature explaining characteristics of
children’s early speech production. When children begin to produce meaningful words, coda
consonants are often omitted (e.g., Altvater-Mackensen & Fikkert, 2010; Demuth & Fee, 1995)
and CV syllables are acquired earlier than CVC syllables in production (Levelt et al., 2000). Recent
experimental evidence and theoretical models claim that children use the same lexical representation
for speech perception and production (Altvater-Mackensen & Fikkert, 2010; Levelt et al., 1999), thus,
the coda omission and later acquisition of CVC syllables as opposed to CV syllables may be the
consequence of children’s better sensitivity to onset over coda position in their lexical representations. Consequently, we may also predict that word-medial onsets will be produced earlier than
word-medial codas in children’s output in a way driven by the asymmetrical nature of their
phonological representations of sounds in these two syllable positions.
The current study is limited in several respects. First, a potential confound which may impact
children’s sensitivity to word-medial onsets and codas is that in word-medial positions, codas are
often undergo regressive assimilation to the place and voicing features of their following consonants.
In contrast, word-medial onsets are less vulnerable to such changes. Thus, it is possible that
children’s lesser sensitivity to coda mispronunciations may due to the fact that codas simply allow
for more changes in the input. Second, as we noted earlier, codas were always in stressed syllables,
while onsets were always in unstressed syllables. Thus, given that onsets were located at the locus of
the change to the second syllable and as well as a change in stress, it is possible that these facts may
make the onset easier to notice than coda, resulting in children’s higher sensitivity to onsets than
458
Y. WANG AND A. SEIDL
codas. Unfortunately, current study is not able to address this issue. It would be thus interesting for
future work to replicate the present study with other types of word forms such that the coda patterns
are in the unstressed syllable and the onset patterns in the stressed syllable in order to examine
possible interactions between syllable position and word stress. In addition, this confound could also
be eliminated by examining trisyllabic or polysyllabic words with the target onset and coda restricted
at the transition of two syllables with relatively equal stress.4
Despite the above-mentioned caveats, we have demonstrated that 24-month-old English-learning
children have an asymmetrical sensitivity to word-medial onsets and codas, indicating a potential
role of phonological biases in children’s word processing. The asymmetrical patterns that we see in
speech perception may also predict children’s production patterns during the course of language
acquisition. In addition, the research provides evidence that by the age of 24 months, children have
very sophisticated phonological knowledge in words such that minimal phonetic deviation in onset
position may affect word recognition.
Notes
1. Although we cannot be sure how infants in our study have interpreted the stimuli, we think it is more likely
that they have interpreted them as bisyllabic trochees as it is a more common pattern in English. With
regard to how children syllabify the stimuli, we think it is more likely they interpreted the stimuli as CVC.
CVC, due to the effect of Weight-Stress Principle and Maximal Onset Principle on syllabification (e.g.,
Hayes, 1995; Kahn, 1976). In addition, most of these consonant-consonant sequences in the word-medial
position, e.g, zm, zg, vd, violate the phonotactic constraints for being a consonant cluster within one syllable
position.
2. However, these differences may be less likely to affect the results. Indeed our results go in the opposite direction
from the acoustic analyses: children show better recognition of target labels in CP trials than in MP trials. We
suspect that the different acoustic measurements between CPs and MPs may due to the fact that the speaker
emphasized the MPs because these were relatively new labels compared to the CPs. Crucially, however, MPonset and MP-coda measures were not distinct prosodically.
3. We also evaluated the possible role of children’s vocabulary size on their representation of segments in wordmedial onset and coda positions. The expressive vocabulary size of these 32 children ranged from 14–100 (M =
55.41, SD = 26.24), as calculated from parental CDI reports. Vocabulary size was strongly correlated with
children’s age, r = 0.453, p = .009. However, we did not find any correlation between children’s vocabulary size
and the effect of naming in any of the Test trials for either PTL measure (/r/s < .165, ps > 0.368) or PLL measure
(/r/s < 0.187, ps > 0.306). These results are consistent with previous findings showing no evidence of a
relationship between vocabulary size and infants’ sensitivity to mispronunciations (Ballem & Plunkett, 2005;
Swingley, 2009).
4. We thank our anonymous reviewers for suggestions related to this discussion.
References
Altvater-Mackensen, N., & Fikkert, P. (2010). The acquisition of the stop-fricative contrast in perception and
production. Lingua, 120(8), 1898–1909. doi:10.1016/j.lingua.2010.02.010
Archer, S., Zamuner, T. S., Engel, K., Fais, L., & Curtin, S. (2016). Infants’ discrimination of consonants: Interplay
between word position and acoustic saliency. Language Learning and Development, 12(1), 60–78.
Bailey, T. M., & Plunkett, K. (2002). Phonological specificity in early words. Cognitive Development, 17(2), 1265–1282.
doi:10.1016/S0885-2014(02)00116-8
Ballem, K. D., & Plunkett, K. (2005). Phonological specificity in children at 1;2. Journal of Child Language, 32(1), 159–
173. doi:doi:10.1017/S0305000904006567
Benavides-Varela, S., & Mehler, J. (2015). Verbal positional memory in 7–olds. Child Development, 86(1), 209–223.
doi:10.1111/cdev.12291
Blevins, J. (1996). The syllable in phonological theory. In J. A. Goldsmith (Ed.), The Handbook of Phonological Theory.
Oxford, UK: Blackwell Publishing.
Capitani, E., Della Sala, S., Logie, R. H., & Spinnler, H. (1992). Recency, primacy, and memory: Reappraising and
standardising the serial position curve. Cortex, 28(3), 315–342. doi:10.1016/S0010-9452(13)80143-8
Clarke, C. M., & Garrett, M. F. (2004). Rapid adaptation to foreign-accented English. The Journal of the Acoustical
Society of America, 116(6), 3647–3658. doi:10.1121/1.1815131
LANGUAGE LEARNING AND DEVELOPMENT
459
Demuth, K. (1995). Markedness and the development of prosodic structure. In J. N. Beckman (Ed.), NELS 25:
Proceedings of the North East Linguistics Society (pp. 13–25). GLSA (Graduate Linguistic Student Association),
Dept. of Linguistics, University of Massachusetts, Amherst, MA.
Demuth, K., & Fee, E. J. (1995). Minimal prosodic words in early phonological development. Unpublished manuscript.
Providence, RI: Brown University; Halifax, Canaca: Dalhousie University.
Denes, P. (1955). Effect of duration on the perception of voicing. The Journal of the Acoustical Society of America,
27(4), 761–764. doi:10.1121/1.1908020
Endress, A. D., & Mehler, J. (2010). Perceptual constraints in phonotactic learning. Journal of Experimental Psychology:
Human Perception and Performance, 36(1), 235–250. doi:10.1037/a0017164
Fennell, C. T., & Werker, J. F. (2003). Early word learners’ ability to access phonetic detail in well-known words.
Language and Speech, 46(2–3), 245–264. doi:10.1177/00238309030460020901
Fenson, L., Pethick, S., Renda, C., Cox, J. L., Dale, P. S., & Reznick, J. S. (2000). Short-form versions of the MacArthur
communicative development inventories. Applied Psycholinguistics, 21(01), 95–116. doi:10.1017/
S0142716400001053
Golinkoff, R. M., Hirsh-Pasek, K., Cauley, K. M., & Gordon, L. (1987). The eyes have it: Lexical and syntactic
comprehension in a new paradigm. Journal of Child Language, 14(01), 23–45. doi:10.1017/S030500090001271X
Goswami, U. C., & Bryant, P. (1990). Phonological skills and learning to read: Essays in developmental psychology. New
York: Psychology Press.
Hayes, B. (1995). Metrical stress theory: principles and case studies. Chicago, IL: Chicago University Press.
Hollich, G. (2008). Supercoder: A program for coding preferential looking (Version 1.7.1). [Computer Software]. West
Lafayette, IN: Purdue University.
Jusczyk, P. W., & Aslin, R. N. (1995). Infants′ Detection of the Sound Patterns of Words in Fluent Speech. Cognitive
Psychology, 29(1), 1–23. doi:10.1006/cogp.1995.1010
Kahn, D. (1976). Syllable-based generalizations in English phonology. (Ph.D.), Massachusetts Institute of Technology,
Massachusetts, MA.
Levelt, C. C. (2012). Perception mirrors production in 14- and 18-month-olds: The case of coda consonants. Cognition,
123(1), 174–179. doi:10.1016/j.cognition.2011.12.001
Levelt, C. C., Schiller, N. O., & Levelt, W. J. (2000). The acquisition of syllable types. Language Acquisition, 8(3),
237–264. doi:10.1207/S15327817LA0803_2
Levelt, W. J., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral Brain
Sciences, 22(1), 1–38. doi:10.1017/S0140525X99001776
Magnuson, J. S., Dixon, J. A., Tanenhaus, M. K., & Aslin, R. N. (2007). The dynamics of lexical competition during
spoken word recognition. Cognitive Science, 31(1), 133–156. doi:10.1080/03640210709336987
Mani, N., & Plunkett, K. (2007). Phonological specificity of vowels and consonants in early lexical representations.
Journal of Memory and Language, 57(2), 252–272. doi:10.1016/j.jml.2007.03.005
Mani, N., & Plunkett, K. (2008). Fourteen-month-olds pay attention to vowels in novel words. Developmental Science,
11(1), 53–59. doi:10.1111/j.1467-7687.2007.00645.x
Mani, N., & Plunkett, K. (2010). Twelve-month-olds know their cups From their Keps and Tups. Infancy, 15(5),
445–470. doi:10.1111/j.1532-7078.2009.00027.x
McMurray, B., Tanenhaus, M. K., Aslin, R. N., & Spivey, M. J. (2003). Probabilistic constraint satisfaction at the
lexical/phonetic interface: Evidence for gradient effects of within-category VOT on lexical access. Journal of
Psycholinguistic Research, 32(1), 77–97. doi:10.1023/A:1021937116271
Miller, J. L., & Eimas, P. D. (1983). Studies on the categorization of speech by infants. Cognition, 13(2), 135–165.
doi:10.1016/0010-0277(83)90020-3
Nazzi, T. (2005). Use of phonetic specificity during the acquisition of new words: Differences between consonants and
vowels. Cognition, 98(1), 13–30. doi:10.1016/j.cognition.2004.10.005
Prince, A., & Smolensky, P. (1993). Optimality theory: Constraint interaction in generative grammar. Boulder, NJ:
Rutgers University, New Brunswick and University of Colorado.
Schmale, R., Cristia, A., & Seidl, A. (2012). Toddlers recognize words in an unfamiliar accent after brief exposure.
Developmental Science, 15(6), 732–738. doi:10.1111/j.1467-7687.2012.01175.x
Swingley, D. (2003). Phonetic detail in the developing lexicon. Language and Speech, 46(2–3), 265–294. doi:10.1177/
00238309030460021001
Swingley, D. (2005). 11-month-old’s knowledge of how familiar words sound. Developmental Science, 8(5), 432–443.
doi:10.1111/j.1467-7687.2005.00432.x
Swingley, D. (2009). Onsets and codas in 1.5-year-olds’ word recognition. Journal of Memory and Language, 60(2),
252–269. doi:10.1016/j.jml.2008.11.003
Swingley, D., & Aslin, R. N. (2000). Spoken word recognition and lexical representation in very young children.
Cognition, 76(2), 147–166. doi:10.1016/S0010-0277(00)00081-0
Swingley, D., & Aslin, R. N. (2002). Lexical neighborhoods and the word-form representations of 14-month-olds.
Psychological Science, 13(5), 480–484. doi:10.1111/1467-9280.00485
460
Y. WANG AND A. SEIDL
Wang, Y., & Seidl, A. (2015). The learnability of phonotactic patterns in onset and coda positions. Language Learning
and Development, 11(1), 1–17. doi:10.1080/15475441.2013.876270
Werker, J. F., Fennell, C. T., Corcoran, K. M., & Stager, C. L. (2002). Infants’ ability to learn phonetically similar
words: effects of age and vocabulary size. Infancy, 3(1), 1–30. doi:10.1207/S15327078IN0301_1
White, K. S., & Morgan, J. L. (2008). Sub-segmental detail in early lexical representations. Journal of Memory and
Language, 59(1), 114–132. doi:10.1016/j.jml.2008.03.001
Zamuner, T. S. (2006). Sensitivity to word-final phonotactics in 9–16-month-old infants. Infancy, 10(1), 77–95.
doi:10.1207/s15327078in1001_5
Zamuner, T. S. (2013). Perceptual evidence for young children’s developing knowledge of phonotactic probabilities.
Language Acquisition, 20(3), 241–253. doi:10.1080/10489223.2013.796951