FACULTEIT DER MAATSCHAPPIJ EN GEDRAGSWETENSCHAPPEN UNIVERSITEIT VAN AMSTERDAM
Datum: 19 maart 2015 Student: Dani Yahiaoui Studentnummer: 6222560
Klinische Ontwikkelingspsychologie
Begeleider(s): Pralle Kriengwatana & Maartje Raijmakers Tweede beoordelaar: Gorka Fraga Gonzalez
Masterthesis
INFANT DISCRIMINATION OF FRENCH
ROUNDED VOWELS
Do infants show perceptual asymmetries when discriminating between the rounded vowel contrasts/ø/- /o/ and /u/-/o/?
1
Index
Abstract………...2
Introduction: Infant Discrimination of Vowels…...…….………3
Method………..10 Participants………...10 Materials………..11 Procedure………..13 Results………....14 Discussion………..18 References………..23
2
Abstract
Research has found that infants show reversed directional asymmetries when comparing rounded vowels. This study investigated the question: Do monolingual Dutch infants show perceptual asymmetries when discriminating between the French rounded vowel contrasts /ø/- /o/ and /u/-/o/? The purpose of this research was to gain more understanding in the underlying processes that play part in infant vowel discrimination and ultimately in understanding how early language development works. The expectation was that infants show directional asymmetries favoring the more central rounded vowel. Unfortunately, the data sample was too small to find conclusive effects. The results did, however, show trends of infants favoring the more central vowel. Further research is necessary in order to say anything conclusive about whether infants do in fact favor the more central vowel. This research provided a theoretical base and a paradigm for future research to start off from. Understanding the underlying processes of language learning may help us in situations where language development goes wrong. If we understand the basis of early language development, we will be able to earlier detect language disorders and develop adequate interventions.
3
Infant Discrimination of Vowels
Infants rapidly learn the sound categories of their native language, and a central goal of language acquisition research is establishing how this ability is acquired during childhood. An important aspect of acquiring a specific language, is learning which phonetic distinctions will be used phonemically in the native language (Kuhl et al. 2006). Various studies have focused on how infants learn these distinctions and found that, in early life, infants are capable of discriminating many, if not all, of the phonetic units of the world’s languages (e.g. Best, McRoberts & Sithole, 1988, Bosch & Sebastián-Gallés, 2003, Kuhl et al., 2006) and, by adulthood, universal phonetic capacity is no longer in place. Nonnative phonetic discrimination can be very difficult for adults (e.g. Best, McRoberts & Goodell, 2001; Iverson, et al., 2003; Miyawaki et al., 1975; Werker & Lalone, 1988; Zhang, Kuhl, Imada, Kotani & Tohkura, 2005, all cited in Kuhl et al., 2006). One of the phonetic units that infants learn to distinguish very early in life are vowels. Infants as young as 6 months have been shown to be sensitive to the vowel categories specific to their native language (Vallabha et al., 2007).
A specific finding in vowel research is that infants show directional asymmetries when discriminating between vowels. This means that when comparing two vowels, infants find a change from one vowel to the other easier than a change in the reverse direction (for example, a change from /e/ to /i/ is easier than from /i/ to /e/, Polka & Bohn, 2003). Polka and Bohn (2003) suggest that the literature provides many examples of vowel asymmetries that, to their knowledge, can only be explained by considering the location of the contrasting vowels within the articulatory/acoustic vowel space (this vowel space is used to represent vowels: F1 has been shown to correlate inversely with the position of the highest part of the tongue in the height dimension, while F2 is correlated with tongue frontness (Watt & Fabricius, 2002), see Figure 1). In particular, infants seem to prefer a change from a more central vowel to a more peripheral vowel, with the peripheral vowel being the vowel more to the edge of the F1/F2 vowel space.
4
Figure 1. Schematised representation of the ‘vowel triangle’ (Watt & Fabricius, 2002)
Polka and Bohn (2003) propose that peripheral vowels (vowels in the periphery or the edge of this F1-F2 vowel space) may serve as important perceptual reference points for human listeners. Several studies considering language development and, in particular, vowel development support this view (Trehub, 1976; Kuhl et al., 1991; Polka & Werker, 1994; Polka & Bohn, 1996; Polka & Bohn, 2003). Polka and Werker (1994) for example found evidence that more familiar (e.g. native) vowels serve as perceptual magnets, and that the bias is thus a language specific bias. Polka and Bohn (1996) examined this by comparing the discriminability of the English /æ/-/ɛ/
contrast and the German /u/-/y/ contrast in 6-8 and 10-12 month-old German and English learning infants. Based on the previous research by Polka and Werker (1994), Polka and Bohn (1996) expected to observe perceptual asymmetries in discrimination of the non-native but not native contrasts. This prediction was not supported. Instead, both English and German infants showed the same perceptual asymmetry for the German /u/-/y/ contrast. With a change from /y/ to /u/ being easier to discriminate than a change from /u/ to /y/. For the English /æ/-/ɛ/
contrast, discrimination was easier if the change occurred from /æ/to /ɛ/ than when it occurred from /ɛ/ to/æ/, irrespective of the age or language background of the infants. So it seems that infants do use particular vowels as reference points, but contrary to what Polka and Werker (1994) proposed, this does not seem to be caused by familiarity with a native language but rather seems to be a language-universal bias (Polka & Bohn, 1996, 2003). This means that the notion of native vowels functioning as perceptual magnets does not serve as an explanation for infant
5
vowel learning. Polka and Bohn (2003) therefore proposed another explanation. After looking at the differences between the vowels used in both studies, Polka and Bohn (2003) found that each of these vowel asymmetries can be accounted for by considering the location within a traditional articulatory or F1/F2 acoustic vowel space as described before (see Figure 2). Figure 2 shows that for the asymmetries that were found, vowel discrimination is easier for infants when they were presented with a change from a less peripheral vowel (more to the centre) to a more peripheral vowel (closer to the limits or corner of the vowel space, see Figure 2). So Polka and Bohn (2003) state that perceptual asymmetries point to a language-universal perceptual bias that infants bring to the task of vowel discrimination such that the relatively more peripheral vowel in a contrast serves as a reference vowel.
Figure 2. Overview of asymmetrical contrasts found in infant research (Polka & Bohn, 2003).
To complement and refine their previous research, Polka and Bohn (2011) suggest that the function of directional asymmetries is to provide a global structure for recognizing acoustic patterns as human speech signals, similar to the early acquisition of a global schema for
recognizing certain visual configurations as a face (e.g. Gliga, Elsabbagh, Andravizou & Johnson, 2009; cited in Polka & Bohn, 2011). They propose the Natural Referent Vowel framework (NRV), which states that vowels with extreme articulatory-acoustic properties (peripheral in
6
vowel space) act as natural referent vowels. Natural referent vowels support and guide the development of vowel perception by attracting infant attention and providing stable perceptual forms for the language learner. Specifically, infants use the vowels: /a/, /u/ and /i/ as referent vowels (within the F1-F2 vowel space these vowels form a triangle in the periphery of the
scheme, see Figure 1) and build from there. The salience and stability of natural referent vowels is due to focalization (results from formant convergence: when formants move close together in frequency there is a mutual reinforcement of acoustic energy that raises the amplitude of each formant; Schwartz, Abry, Boe, Ménard & Valée 2005; Polka & Bohn, 2011). Furthermore, the NRV framework states that experience with native vowels will cause the directional asymmetry to disappear over time, asymmetries will only be maintained if listeners lack experience with specific vowels (e.g. non-native vowels, Polka & Bohn, 2003).
The NRV complements several prominent models of language acquisition that have been proposed to explain vowel development and, in particular, directional asymmetries. Kuhl (2000), for example, offered a model of early speech perception termed the Native Language Magnet (NLM) model. This model accounts for the language-universal bias and defines three stages in infants’ language development. In the first stage, infants are able of discriminating between all of the world’s phonetic units. Kuhl (1991) states that this understanding originates from infants’ general auditory processing mechanisms rather than from a speech-specific mechanism. In the second stage, infants gain experience, causing a bias that increases perceptual sensitivity near the boundaries between categories and decreases perceptual sensitivity near category modes (Kuhl, 1991; Kuhl et al., 1993; Iverson et al., 2003). As the experience of infants increases, the most activated representations start functioning as perceptual magnets for other members of the category, thus increasing the perceived similarity between members of that category (Kuhl, 1991). In stage 3, this bias of perception, described as the perceptual magnet effect, facilitates native language phonetic abilities and reduces non-native language phonetic abilities (Kuhl et al., 2008).
7
The NLM (Kuhl, 1993; Kuhl et al., 2008) provides a basis for understanding how
experience shapes phonetic categories. To provide a more comprehensive framework, PRIMIR (a developmental framework for Processing Rich Information from Multidimensional Interactive Representations) was proposed, pulling together the processing and the storage of information, which fundamentally depend on the age of the infants, the nature of the task, and the infant’s perceptual biases (Curtin, Byers-Heinlein & Werker, 2011). PRIMIR assumes there is rich information available in the speech input and that the child picks up and organizes this
information along a number of multidimensional planes. Use of this rich information depends on the joint activity of three dynamic filters. These filters – the initial biases, the developmental level of the child and requirements of the specific language task the child is facing – work together to differentially direct attention to 1 (or more) plane (Werker & Curtin, 2005).
Thus, directional vowel asymmetries can be seen as initial biases that fit the developmental level of an infant and help the infant to accomplish a specific language-learning task. And,
according to the above mentioned studies it seems that the F1/F2 vowel space provides us with a clear cut explanation as to why directional asymmetries occur.
However, studies have found that not all directional asymmetries can be explained by the relative location of vowels in the F1/F2 vowel space. For vowel contrasts involving lip-rounding, the vowel that is easier to discriminate within a pair, may not necessary fall closer to the edge or corner of the F1/F2 vowel space. Two studies have found directional asymmetries in the opposite direction (i.e. peripheral to central was easier to discriminate; Best & Faber, 2000, cited in Polka & Bohn, 2003; Polka & Bohn, 2011). Both of these studies used vowel contrasts that involved one vowel that required lip-rounding and another that did not. Polka and Bohn (2011) found that, when infants (6- to 12-months-old) compared the Danish vowels /e/ and /ø/, the directional asymmetry did not favor /e/, which is more peripheral than /ø/ in the F1/F2
acoustic space. A similar result was found by Best and Faber (2000; cited in Polka & Bohn, 2003). They found that the youngest infants (3-5 months-old) showed directional asymmetries towards
8
the more central vowel, when tested with the native Norwegian contrast /i/-/y/ (with /y/ as the rounded vowel). Researchers de Jonge and Boersma (unpublished) found the same unexpected directional asymmetry in French adults. In this experiment, recordings of brain activity were made while participants listened to four different French vowels: /y/, /u/, /o/ and /ø/. The research design was a 3 deviant-oddball task, so one standard vowel sound was repeated and at three random moments one of the other vowels was played (e.g. y, y, y, o, y, y, u, y, y, y, y). According to the NRV framework, the expectation would be that the participants would show directional asymmetries toward the more peripheral vowel (i.e. /ø/ /u/; /y/ /u/; /o/ /u/; /ø/ /o/; /ø/ /y/ would elicit greater brain responses). However, the results showed the opposite pattern; higher brain activity when the vowel switches from peripheral to more central vowels (de Jonge & Boersma, unpublished). This result is unexpected because research shows that in mature perceivers vowel perception biases will be maintained or enhanced for non-native vowel contrasts, but will be reduced or absent for non-native vowel contrasts in many phonetic tasks (Polka & Bohn, 2003, 2011). Since the vowels used in the research were native, it is
remarkable that the directional asymmetries show a reversed pattern.
The above mentioned results may be caused by the contrast between a rounded and an unrounded vowel, or may be due to properties of rounded vowels themselves (i.e. they do not follow the pattern of unrounded vowels in F1/F2 vowel space). It is interesting to further explore this phenomenon in order to gain more knowledge about why the location in the F1/F2 vowel space does not seem to provide an explanation for rounded vowels. Since the foundation of vowel learning is laid in infancy, it is interesting to see if infants also show the same reversed pattern as adults when discriminating between these French rounded vowels. If this result is replicable in infants, it seems to implicate that rounded vowels are processed in a different way than unrounded vowels. And that will mean that further research is necessary to uncover these underlying processes.
9
As mentioned earlier, the NRV framework states that experience with native vowels will cause directional asymmetries to disappear over time, and directional asymmetries will only be maintained for non-native vowels (Polka & Bohn, 2003). Polka & Bohn (2011) found, that for the youngest Danish-learning infants (6;0 – 7;16 months old) a change from less peripheral (/ø/) to more peripheral (/e/) was significantly less discriminable than a change from /e/ to /ø/ for native vowels. For the older Danish-learning infants (7;22 – 11;9 months) direction of
presentation did not significantly affect the discriminability from /ø/ to /e/. This finding, that older infants do not show a discrimination in any particular direction, corroborates with the NRV framework. That the brains of adults also respond more to central vowels than peripheral vowels suggests that this reversed result is due to developmental changes or that the pattern of vowel asymmetries is not universal to all vowels.
Consequently this research will focus on examining Dutch infants’ discrimination of two French rounded vowel contrasts: /ø/ - /o/ and /u/ - /o/. The same French vowels (non-native for Dutch infants) as in the previously mentioned adult study (de Jonge & Boersma, unpublished) will be used, in order to compare the adult results with infant research. Additionally, directional asymmetries in perception of these vowels have not yet been determined in infant research. Therefore, this research addresses outstanding aspects of the NRV by asking the following question: Do infants show perceptual asymmetries when discriminating between the rounded vowel contrasts/ø/- /o/ and /u/-/o/?
To examine this question the vowel contrasts /u/-/o/ and /ø/-/o/ are tested in infant vowel learning. The expectation is that infants will show asymmetries during discrimination of /u/-/o/ and during discrimination of /ø/-/o/. Considering that studies have found that a change from a peripheral to a central vowel is easier to discriminate if one of the vowels is rounded (Polka & Bohn, 2003, 2011; de Jonge & Boersma, unpublished), we predict the same when the contrast involves two rounded vowels. So, instead of favoring the more peripheral vowel, infants will favor the more central vowel; in this case respectively /u/ and /ø/. If infants find a change from
10
a peripheral to a central vowel easier to discriminate, this could suggest that rounded vowels are processed differently than unrounded vowels. If infants find a change from a central to a peripheral vowel easier to discriminate, this means that directional asymmetries are likely to be explained by the NRV framework.
Method
Participants
22 infants participated in the experiment, 2 participants had to be excluded: they were unable to complete the minimal amount of six familiarization trials due to fuzziness. So 20 participants, between the ages of 10,1 and 11,3 months-old, completed this experiment.
Participants were selected from a pre-existing database consisting of infants between the ages of 4 to 18 months-old. Participants were only invited if they were raised monolingually (that is, if they heard Dutch at least 75% of the time). Participants were randomly assigned to one of the eight conditions (see Table 1). As a reward for participating in the research, participants received
a children’s book.
Table 1
Overview of the vowel contrasts compared in each condition
Condition Familiarization trials Test 1 trials Test 2 trials
1 o o u 2 o u o 3 u o u 4 u u o 5 o o ø 6 o ø o 7 ø o ø 8 ø ø o
11 Materials
The present experiment is similar to the task-switching paradigm which requires people to switch rapidly between tasks (Jersild, 1927, cited in Kramer, Hahn & Gopher, 1999), however it is modified so that infants will be able to perform on the task. Infants are first familiarized (pre-trained) with the task and the stimulus (in this case one of the vowel sounds /o/, /u/ or /ø/). After the infant is familiarized with the vowel sound he or she will hear two test trails. In one of the test trials the same sound as in the familiarization trials is repeatedly played and in the other test trial a different sound is repeatedly played, these test trials will be counterbalanced (see Table 1). This will require the infants to remember the vowel sounds that they heard in the
familiarization sounds and simultaneously realize that this sound is either different or the same. General language development of the infants is measured with the general language development questionnaire. This questionnaire consists of 15 items, examples of questions are: ‘Does your child have or ever had hearing problems?’, ‘Which language is spoken at home?’, ‘Which language is spoken by parents and close family?’ and ‘Does any of your relatives have language disorders?’. This questionnaire is mainly used to check for abnormalities (e.g. problems with hearing or with understanding language) and to make sure that infants are raised
monolingually.
Communicative development is measured with the N-CDI 1 short version (the
Communicative Developmental Inventory). This is a short form suitable for infants from 8 to 16 months-old, the questionnaire is comprised of 206 items with separate columns for word
comprehension (103 items) and word production (103 items). The parents are asked to fill out which words and sentences they think their infant understands. Depending on the age of the infant the parents will be asked to also complete the production part (if a child is not yet producing words, this section will be left out). This originally English test is adapted for Dutch and Belgian infants, it has been validated for those languages by Zink and Lejaegere (2003). The reliability of word comprehension is (>.98) and the reliability of word production is (>.97). So
12
both questionnaires are very reliable, with a comprehension validity of r =.62. The original test, the CDI: Words and Gestures (Infant form) has been produced by MacArthur-Bates (http://mb-cdi.stanford.edu/).
The French vowel sounds /u/, /o/ and /ø/ spoken by a male were the same vowel sounds as used in the study by de Jonge and Boersma (unpublished), who used the synthetic values from Fougeron and Smith (2003). Figure 3 shows the location of the vowels in the F1/F2 vowel space, with /o/ being the more peripheral vowel in both contrasts.
Figure 3. Location of vowels /o/, /u/ and /ø/ in the F1/F2 vowel space
Apparatus
A Tobii 1750 eye-tracker with a 50 Hz sampling frequency (20 ms per sample) was used to capture where the infants were looking on the screen, in order to gather reliable data on whether or not infants heard a change in vowel sounds. E-prime (Psychology Software Tools, Sharpsburg, PA, USA) was used for task control and data collection and point of gaze was calibrated through the native Clearview software. Trials were shown on the Tobii monitor and two speakers located at the infants’ eye level played sound. Trial number, x and y coordinates of the upper left corner of the stimulus, x and y coordinates of the infants’ gaze and timing were collected (Schure et al., 2014). 600 700 800 900 1000 1100 1200 1300 1400 1500 200 250 300 350 400 F1 ( He rt z) F2 (Hertz) French eu French u French o
13 Procedure
First the parent(s) of the infant had to fill out a questionnaire regarding general
development. This questionnaire consisted of 15 questions, examples of questions were: ‘Were there any complications during pregnancy?’, ‘Has your child ever had an ear infection?’ and ‘Which languages does your child hear at home?’. After completing the questionnaire and signing the consent form, the infant and parent(s) were taken into another room, where the infant was seated in a baby seat in front of the Tobii eye-tracker. The monitor of the eye-tracker was adjusted so that the infant’s eyes were clearly visible for the Tobii-camera. When the infant was properly seated, the parent was seated behind the baby seat, so that he or she did not disturb the experiment but was still close enough in case the infant needed comforting. The researcher went into another room with a one-way mirror during the experiment.
The two vowel contrasts /o/-/ø/ and /u/ - /o/ were counterbalanced, so there were 8 conditions in total. Participants were randomly divided into these eight conditions. Each condition consisted of six familiarization trials and two test trials. In one of the test trials the same sound as in the familiarization trials was played repeatedly and in the other test trial a different sound was played repeatedly. For example, condition 1 consisted of: 6 familiarization trials repeating the vowel sound ‘o’, test trial 1 repeating the same sound (vowel ‘o’) and test trial 2 repeating a different sound, in this case vowel sound ‘u’ (see Table 1). While listening to these vowel sounds a target was shown on the screen (see Figure 4). Between trials the infant saw an attention getter (a moving picture that was appealing for the infant) to attract the infant’s attention.
The eye-movements and, in particular, the looking time of the infant were measured with the Tobii eye-tracker. After the experiment was completed, the researcher took the infant and parent(s) back into the other room, where the parent was asked to fill out a shortened version of the N-CDI. This is a questionnaire consisting of 103 items regarding ‘early word comprehension’. If infants were already producing words, an additional 103 item questionnaire was used to
14
measure ‘word production’. As an acknowledgement for their help, participants received a children’s book.
Figure 4: Target
Results
Of the 22 participants tested, 2 participants were excluded from the test due to fuzziness. The remaining 20 participants were equally divided by sex, 10 male and 10 female. A power analysis - calculated using G*power (calculated effect size F=.02, α error probability=.05, and 8 groups) – showed that the data sample was too small to find conclusive effects (power of .051). The NCDI and the general language questionnaire were used to check for irregularities within the sample. When reviewing the answers the results showed large standard deviations compared to the mean. It seemed that infant word production and word comprehension differed quite a lot in this relatively small age group, shown in Table 2.
Table 2
Mean and Standard deviation scores of Word Comprehension and Word Production for All 20 Participants
Mean Std. Deviation
Word comprehension 45.75 31.05 Word production 61.50 22.82
By means of a 2x2 repeated measures ANOVA, we first investigated whether infants show directional asymmetries when discriminating between rounded vowel contrasts. The dependent variable was looking time, the within-subject independent variables were (1) change and (2) no change: change means that infants listened to a test trial where the vowel sound
15
changed to a different sound than the familiarization trials and no change means that participants hear the same vowel sound as in the familiarization trials. Change direction (either from
peripheral to central, or central to peripheral) was the between subjects factor. No significant effect was found for the test (F(1,18)=.02, p>.05), which means that infants did not show a difference in looking time whether they heard a vowel change or no vowel change. The direction of the vowel change also did not influence the looking time in either a change or a no-change trial (F(1,18)=.74, p>.05). To check if the age of the participant affected the results, age was added as a covariate. Still no significant effect was found for the test, infants did not show a difference in looking time for either a change or no change trial regardless of age (F(1,17)=2.30, p>.05). The direction of the vowel change did not influence looking time (F(1,17)=1.16, p>.05) and age did not seem to influence this effect either: F(1,17)=1.04, p>.05.
This broad analysis did not account for the different vowel contrasts that were tested. Perhaps the participants found one of these contrasts easier to discriminate than the other. To account for these vowel contrast differences, the data was split by contrast (so the participants that listened to the /ø/-/o/ vowel contrast were analyzed separately from the participants that listened to the /u/-/o/vowel contrast). When age was not included as a covariate, no significant results were found for either of the contrasts. Infants did not look differently at a change or no-change trial (/ø/-/o/ contrast F(1,7)=.740, p>.05, /u/-/ø/ contrast F(1,9)=1.27, p>.05) nor did the direction of the vowel change influence the looking time at a change or no-change (/ø/-/o/ contrast F(1,7)=.010, p>.05, /u/-/o/ contrast F(1,9)=2.46, p>.05).
Since research has found that older infants show an increase in directional asymmetries for non-native vowel contrasts (Polka & Bohn, 2003), it could be that even within our small age group age had an influence. So age was added as a covariate. When accounted for age in the split data file, significant results were found for the /ø/-/o/ contrast. When listening to the /ø/-/o/ contrast infants show a difference in looking time between change and no change trials. So they seem to look more when the vowel sound changes than when it stays the same (F(1,6)=15.92,
16
p<.05, partial η2= .73). Furthermore, age seems to influence whether infants look more at a vowel
change than at a no-change trial (F(1,6)=16.33, p<.05, partial η2=.73). Looking at Figure 5, it
seems that older infants tend to look more when a vowel changes than when it does not change. No significant effect was found for the influence of a change in direction (peripheral or central) on looking time for change versus no change trials F(1,6)=2.57, p>.05. However, when looking at the means adjusted for age, there seems to be a trend. Infants seem to prefer a vowel change from a peripheral to a more central vowel (see Table 3). Interestingly, no significant results were found for the /u/-/o/ contrast: When comparing change versus no-change trials infants do not show a difference in looking time: F(1,8)=.29, p>.05. Age does not seem to influence amount of looking when comparing change versus no-change: F(1,8)=.33, p>.05. And no effect was found for the influence of a change in direction on looking time: F(1,8)=2.2, p>.05. When looking at the means in Table 3, participants seem to have a slight preference for a change from peripheral to central, but since the data is not significant no conclusive statements can be made about this trend.
Figure 5
Differences in individual looking times (compared to the average of the last two familiarization trials), split by trial type (change and no change) and by contrast type (/ø/-/o/ and /u/-/o/)
17 Table 3
Amount of Looking at a Test Trial (Either Change or no Change) Compared to the Average of the Last Two
Familiarization Trials, with (Standard Errors), Split by Change Direction (to Central or to Peripheral) for Both Contrasts (/ø/-/o/ and /u/-/o/)
Change direction Mean (std. error)
95% Confidence interval
Lower Bound Upper bound
/ø/-/o/ contrast To central .985 (.328) To peripheral .444a (.370) .183 1.787 -.462 1.350 /u/-/o/ contrast To central -.280 (1.173) -2.986 2.426 To peripheral -.152b (1.285) -3.116 2.813
a. Covariates are evaluated at the following values: age = 10.63 b. Covariates are evaluated at the following values: age =10.55
The average looking time of the familiarization trials was plotted for the three vowel sounds. Figure 6 illustrates that infants seem to familiarize the vowels sounds, they look less at the screen the longer they hear the vowel sound. What is interesting is that for the vowels /ø/ and /u/ the amount of looking seems to increase between the 5th and 6th trial. However a
paired-sample t-test showed that these differences are not significant (Table 4). The fact that infants seem to look less at the familiarization trials over time means that the experiment works and infants are getting used to the vowel sound.
Figure 6
Average looking time of the 6 familiarization trials for all three vowel contrasts
0 5000 10000 15000 20000 1 2 3 4 5 6 Lo ok in g t im e Familiarization trials
Looking time over familiarization
trials
mean eu mean u mean o
18 Table 4
Mean Difference Between the 5th and 6th Familiarization Trials for Each Vowel Sound, with Standard Deviation and
Significance Level.
95% Confidence Interval
5th - 6th triala Mean (Std. Dev.) Lower Upper Sig. (2-tailed)
/o/ -476.3 (1491.2) -1543,1 590,5 .339
/u/ -1992.8 (2968.1) -5107,7 1122 .161
/ø/ -1430.5 (1487.6) -3797,6 936,6 .150
aDifference between the 5th and 6th familiarization trial for each vowel sound.
Discussion
So, interestingly enough, the only significant effects that were found considered the /ø/-/o/ contrast. When discriminating between /ø/and /ø/-/o/, infants were able to hear a difference between a vowel change and no change and age seemed to influence whether the infants looked more at a vowel change than at no vowel change. In particular, older infants seemed to look more at a vowel change than younger infants. This is in accordance with previous research that has found that directional asymmetries slowly disappear for native contrasts across the first year of life but are enhanced for non-native contrasts (Polka & Bohn, 2003). And since the used French vowels were non-native contrasts for Dutch infants it explains why older infants tend to look longer.
Unfortunately, due to the small data sample, it is not possible to say anything conclusive about the research question. There seems to be evidence that infants show directional
asymmetries when comparing rounded vowels, and specifically, there seems to be a trend towards the more central vowel. Furthermore, when we compare our research results with the results of the adult study of de Jonge and Boersma (unpublished), we see some similarities. They found a significant effect for the /ø/-/o/ contrast (p=.003), but a trend for the /u/-/o/contrast
19
the /u/-/o/ contrast. It seems that in both our research as in the French adults research the vowels in the /ø/-/o/ contrast are easier to discriminate between than in the /u/-/o/ contrast. So, although nothing conclusive can be said about the direction that infants favor most when comparing rounded vowels, there seems to be enough reason to continue testing with a larger sample size. There is still an indication that there are development differences in the learning of rounded vowels that are not yet predicted by the NRV framework. This would mean that the natural referent vowel framework as proposed by Polka & Bohn (2003, 2011) does not provide an explanation for all vowel learning.
As mentioned above, an interesting fact is that we only found significant data for the /ø/-/o/ contrast. A possible explanation for this could be that the /ø/-/ø/-/o/ contrast is more clearly distinguishable and thus it is easier for infants to discriminate between the vowels. The /u/-/o/ vowels seem to sound more similar and may therefore be harder to discriminate between. When looking at their relative physical location in the vowel space, this could be the case. As shown in Figure 7, the Dutch vowels /u/ and /o/ are further apart, while the French /u/ and /o/ vowels are quite close in this physical space. This could serve as an explanation to why Dutch infants find it more difficult to discriminate between the French /u/-/o/ contrast, and why no effect was found for this contrast. This is something further research should take into account. Maybe a comparison of both French and Dutch vowels could be made, in order to see if infants react differently to these contrasts and to check how large the differences between Dutch and French vowels are.
20 Figure 7.
The Relative Location of French a and Dutch b Vowels /o/, /u/ and /ø/ in the F1/F2 Vowel Space
a Values for adult male French stimuli are taken from Fougeron and Smith (1993)
b Values for adult male Dutch stimuli are taken from Adank et al. (2004)
Besides testing a larger sample and adding Dutch vowels, it might be interesting to add a younger age group. The age group that was tested in this experiment was between the ages of 10,1 and 11,3 months-old. This is not a very large age difference, and more importantly it is around the age were infants normally learn different strategies to cope with discrimination (Kuhl et al., 2006). It would be interesting to test a younger age group (8 or 9-month-olds for example) alongside the older age group, to see if they respond differently to the vowel contrasts. We would expect younger infants to show larger directional asymmetries when tested for native vowels, but smaller asymmetries when tested for nonnative vowels.
Another finding was that large variances were found within the NCDI scores of Word
Comprehension and Word Production. This has perhaps something to do with the fact that it is a parent questionnaire, and some parents reported that they found it quite difficult to fill in
whether or not their child understood words or sentences. Since it is quite subjective, it is logical that quite large differences between participants occur.
One thing that was surprising in the sample group was that at least 6 of the infants had been hospitalized right after birth for at least one day with a maximum of five days in this subset. Douglas (1975) found evidence that early hospital admission for up to a week is not associated
600 800 1000 1200 1400 1600 1800 200 250 300 350 400 450 F1 ( He rt z) F2 (Hertz) Dutch ø Dutch u Dutch o French ø French u French o
21
with increased risk of later behavioral or emotional disturbance. Hospital admission of more than a week’s duration or repeated admission before the age of five, however, is associated with an increased risk of behavioral disturbance and poor reading in adolescence (Douglas, 1975). Since no hospital admission longer than five days was reported for the subset, we do not expect hospital admission to have an influence on infant performance.
In conclusion, we can state that despite the low power and few significant results, the trends and resemblance with de Jonge and Boersma’s (unpublished) study offers enough reason to continue testing. Rounded vowels seem to act differently than unrounded vowels and this
indicates development differences in the learning of rounded vowels that are not yet predicted by the NRV framework.
Perhaps the NLM framework (proposed by Kuhl, 1994, 2000) provides an explanation for rounded vowels, in that – with experience – the representations that are most often activated (prototypes) begin to function as perceptual magnets for other members of the category. This would mean that the learning of rounded vowels will differ between different languages, and thus is not language-universal. Both the studies of Best and Faber (2000, cited in Polka & Bohn, 2011) and Polka and Bohn (2003) only tested native rounded vowel contrasts and have not tested non-native vowels. The adult study of de Jonge & Boersma also focused on non-native vowel contrasts. In spite of finding trends in the current research that seem to provide evidence that infants do in fact show a universal-language bias for rounded vowels, it is not significant, so further research is necessary to provide a conclusive answer. This provides another reason to test both Dutch and French vowels, in order to compare infant reactions to those vowels. If the bias for rounded vowels is language-universal, we would expect infants to show an increase in directional
asymmetries for non-native vowels, as they get older and a decrease in native vowels. If the bias is not language-universal, we would expect infants to show directional asymmetries for the native vowels, but not for the non-native vowels.
22
Although the research question whether infants show directional asymmetries when comparing French rounded vowels could not be conclusively answered, this research is relevant in that it provides a clearer understanding in vowel discrimination research in infants. The used paradigm has proven to work and thus provides a paradigm to continue the testing of rounded vowel learning in infants. It is important to continue research, because if we have a clearer understanding of the processing of vowels in infancy we will be able to better understand early language development in humans and hopefully be able to treat language disorders or situations where language development goes wrong. This will – in the end – lead to better care for children and adults with language disorders.
23
References
Adank, P., Van Hout, R., & Smits, R. (2004). An acoustic description of the vowels of Northern and Southern Standard Dutch. The Journal of the Acoustical Society of America, 116(3), 1729-1738.
Best, C. T., Faber, A. (2000). Developmental increase in infants’ discrimination of non-
native vowels that adults assimilate to a single native vowel. In: International Conference on Infant Studies, Brighton, UK, 16-19 July 2000.
Best, C. T., McRoberts, G. W., & Sithole, N. M. (1988). Examination of perceptual
reorganization for non-native speech contrasts: Zulu click discrimination by English-speaking adults and infants. Journal of Experimental Psychology: Human Perception and Performance, 14, 345-360.
Bosch, L., & Sebastián-Gallés, N. (2003). Simultaneous bilingualism and the perception of a language-specific vowel contrast in the first year of life. Language and Speech, 46, 217-243. Curtin, S., Byers-Heinlein, K., & Werker, J. F. (2011). Bilingual beginnings as a lens for
theory development: PRIMIR in focus. Journal of Phonetics, 39(4), 492-504. De Jonge, M. & Boersma, P., unpublished.
Douglas, W. B. (1975). Early hospital admissions and later disturbances of behaviour and learning. Developmental Medicine & Child Neurology, 17(4), 456-480.
Fougeron, C., & Smith, C., 1993. ‘French’, Journal of the International Phonetic Association, 23, 73-76. Iverson, P., Kuhl, P. K., Akahane-Yamada, R., Diesch, E., Tohkura, Y., Kettermann. A.,
& Siebert, C. (2003). A perceptual interference account of acquisition difficulties for non-native phonemes. Cognition 87(1), B47-B57.
Kramer, A. F., Hahn, S., & Gopher, D. (1999). Task coordination and aging: Explorations of executive control processes in the task switching paradigm. Acta Psychologica, 101(2), 339-378.
24
for the prototypes of speech categories, monkeys do not. Perception and psychophysics, 50, 93-107.
Kuhl, P. K. (1993). Early linguistic experience and phonetic perception: implications for theories of developmental speech perception. Journal of Phonetics 21, 125–139. Kuhl, P. K. (1994). Learning and representation in speech and language. Current Opinion in
Neurobiology, 4(6), 812-822.
Kuhl, P. K. (2000). A new view of language acquisition. Proceedings of the National Academy of Sciences, 97, 11850-11857.
Kuhl, P. K., Conboy, B. T., Coffrey-Corina, S., Padden, D., Rivera-Gaxiola, M., &
Nelson, T. (2008). Phonetic learning as a pathway to language: new data and native language magnet theory expanded (NLM-e). Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1493), 979-1000.
Kuhl, P. K., Stevens, E., Hayashi, A., Deguchi, T., Kiritani, S., & Iverson, P. (2006).
Infants show a facilitation effect for native language phonetic perception between 6 and 12 months. Developmental Science, 9(2), F13-F21.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992). Linguistic experience alters phonetic perception in infants by 6 months of age. Science, 255(5044), 606-608.
MacArthur-Bates Communicative Development Inventories (2007). Acquired on 17 November 2014 from http://mb-cdi.stanford.edu/.
Polka, L., & Bohn, O. S. (1996). A cross-language comparison of vowel perception in English- learning infants and German-learning infants. The Journal of the Acoustical Society of America, 100, 577-595.
Polka, L., & Bohn, O. S. (2003). Asymmetries in vowel perception. Speech Communication, 41(1), 221-231.
25
emerging view of early phonetic development. Journal of Phonetics, 39(4), 467-478. Polka, L., & Werker, J. F. (1994). Developmental changes in perception of nonnative
vowel contrasts. Journal of Experimental Psychology: Human Perception and Performance, 20(2), 421.
Schure, S., Mandell, D. J., Escudero, P., Raijmakers, M. E., & Johnson, S. P. (2014). Learning stimulus-location associations in 8- and 11-month-old infants: multimodal versus unimodal information. Infancy, 19(5), 476-495.
Schwartz, J. L., Abry, C., Boë, L. J., Ménard, L., & Vallée, N. (2005). Asymmetries in
vowel perception, in the context of the Dispersion–Focalisation Theory. Speech Communication, 45(4), 425-434.
Trehub, S. E. (1976). The discrimination of foreign speech contrasts by infants and adults. Child Development, 466-472.
Vallabha, G. K., McClelland, J. L., Pons, F., Werker, J. F., & Amano, S. (2007). Unsupervised learning of vowel categories from infant-directed speech. Proceedings of the National Academy of Sciences, 104(33), 13273-13278.
Watt, D., & Fabricius, A. (2002). Evaluation of a technique for improving the mapping
of multiple speakers’ vowel spaces in the F1-F2 plane. Leeds Working Papers in Linguistics and Phonetics, 9(9), 159-173.
Werker, J. F., & Curtin, S. (2005). PRIMIR: A developmental framework of infant speech processing. Language Learning and Development, 1(2), 197-234.
Zink, I., & Lejaegere, M. (2003). N-CDIs: Short forms, adaptation and norms of the Dutch version of the MacArthur Short Form Vocabulary Checklists of Fenson et al. Leuven, Belgium: Acco.
