Beneficial Effects of the Mother’s Voice on Infants’ Novel Word-Learning and its Relationship with Vocabulary Size and Gaze Following Abilities
Eline L. Bekkers Utrecht University
Abstract
Building a vocabulary starts with learning associations between words and objects. Large variation is observed in this development. It is therefore interesting to determine which factors facilitate novel word learning. Language studies in infants take special interest in the
influence of the mother‘s voice on language processing. However, no direct evidence for a beneficial effect of the mother‘s voice on novel word learning has been found. In this eye-tracking study, we determined whether 24-month-olds learnt new word-object associations more easily from their mother than from a stranger. At the individual level, we tried to explain some of the variation in word-learning performance by looking at correlations with both productive vocabulary size and gaze following behaviour. We provide the first direct evidence for a facilitating effect of the mother‘s voice on children‘s word-object mapping. Moreover, we found that variation in word-learning ability positively correlates with current productive vocabulary size, supporting the notion that the vocabulary spurt is the result of the
development of more efficient word-learning strategies. We did not find a direct link between gaze following abilities and word-learning skill, nor did we provide evidence for the notion that children are more sensitive to social cues from their mother compared to a stranger.
Introduction
A crucial part of language development is building a vocabulary. In order to do this, children must be able to make associative links between words and their referents by realising that a certain word co-occurs with a certain object (Werker, Cohen, Lloyd, Casasola & Stager, 1998). Forming such word-object associations is not a trivial task, due to the highly variable context in which both words and their referents are presented. Nonetheless, children are able to learn new word-object associations with remarkable speed and ease. For instance, 13- and 18-month-olds learn new word-object associations after just nine times of exposure in a 5 minute training period (Woodward, Markman & Fitzsimmons, 1994). By the time they are 18 to 24 months of age, children‘s vocabulary starts to increase rapidly, resulting in a productive vocabulary of about 50 words, which grows quickly to about 500 words in the third year (Hoff, 2013).
It is important to note, though, that there exists a great variability in these figures across individuals. For instance, an analysis of CDI vocabulary measures showed that word comprehension at 16 months spanned from 92 to 321 words in the 10th and 90th percentile respectively. Although the overall correlation between comprehension vocabulary and age was positive (.60, p < .001), age only accounted for 36% of the variation (Fenson et al., 1994). This great variation in the overall trend of vocabulary development raises the question why some children are able to form word-object associations more easily than others. Knowing more about the facilitating factors for word-object mapping may lead to more insight into what causes some children‘s vocabulary development to fall below the average. Therefore, it can be of great use for developing treatments for language delays.
Some factors facilitating word-object mapping have been identified in previous research. For instance, the way in which parents or caretakers speak to their children influences how well the child learns a new word-object association. A very important
example is the usage of child-directed speech. This way of speaking, characterised by for instance slower rate, longer vowels and pauses, and increased repetition (e.g. Fernald & Simon, 1984; Ratner & Luberoff, 1984; van de Weijer, 1997), is often automatically used by adults when they are speaking to an infant. Its effect on children‘s ability to link words to objects was investigated in an eye-tracking study with 21-month-olds (Ma, Golinkoff, Houston & Hirsh-Pasek, 2011). Children were shown animations of novel objects accompanied by a short story introducing the object‘s name in either adult- or child-directed speech. During the test phase, objects were shown side-by-side, and infants were directed to look at one of the objects. The relative looking times at the named and unnamed object were then analysed. The study showed that 21-month-olds only learnt the new word-object associations in the child-directed speech condition.
Apart from such verbal influences, social factors also seem to facilitate novel word-learning. Several researchers have studied the relationship between joint-attention and vocabulary size. For instance, children comprehended newly learnt words better when their mother had used the new word to indicate an object that the child was already attending to, rather than an unattended object (Dunham, Dunham & Curwin, 1993). Moreover, observational studies showed that the number of joint-attention episodes between mother and child at the age of 15 months was positively related to the child‘s vocabulary at 21 months (Tomasello & Farrar, 1986).
Speaker familiarity is another effect that has been extensively studied in infant language research. In particular, there seems to be a special role for the mother‘s voice in infants‘ language processing. Children generally have the most experience with their mother‘s voice compared to any other voice. In fact, even before birth, foetuses show differential behavioural responses in reaction to their mother‘s voice compared to an unfamiliar voice, indicating that they become familiar their mother‘s voice at a very early stage through
frequent in-utero exposure (Kisilevsky et al., 2003). Moreover, infants younger than 3 days of age have been found to actively adapt their sucking behaviour in order to hear their mother‘s voice instead of a stranger‘s voice (DeCasper & Fifer, 1980). With regard to speech processing, an electrophysiological study which compared the activity patterns elicited by a vowel stimulus pronounced by either the mother or a stranger showed that the former voice was predominantly processed in the left temporal lobe, an area highly associated with language processing. In contrast, the stranger‘s voice elicited a response predominantly in the right temporal lobe (Beauchemin et al., 2010). These results suggest that infants‘ voice-processing brain areas are functionally tuned for familiar voices. Similar outcomes were found in a neuroimaging study which showed significantly stronger responses in the posterior temporal region for the mother‘s voice compared to a stranger‘s voice (Dehaene-Lambertz et al., 2010).
In light of these findings, the mother‘s voice is an interesting factor to study in child language tasks. So far, most studies in this field have focused on its influence on children‘s processing of familiar words. An ERP study looking at congruous and incongruous object-label pairs showed that 9-month-old infants detect the mismatch in the incongruous pairs only when their mother introduces the label (Parise & Csibra, 2012). This finding is supported by the fact that evidence of word comprehension at very young ages (6 8 months) is only found in studies which use the mother‘s voice (although there is no direct comparison with an unfamiliar voice in these studies, e.g. Bergelson & Swingley, 2012). With respect to learning novel words, one study has shown that children are better able to recognise new word forms (segmented speech fragments without meaning) when they are taught by their mother compared to a stranger (Barker & Newman, 2004).
These studies show that the mother‘s voice can facilitate the processing of already familiar words by young infants. It is, however, unclear whether the mother‘s voice also has
an effect on how new word-object associations are learnt. There are several theories about the possible beneficial effect of the mother‘s voice for novel word-learning. A social account states that the mother‘s voice may help children to understand the intention of the word-learning situation better, because they are more sensitive to social cues coming from their mother versus a stranger (e.g. Bruner, 1981; Smith, 2000; Spelke, Bernier & Skerry, 2013; Tomasello, 2003, but see Pruden, Hirsh-Pasek, Golinkoff & Hennon, 2006). An acoustic account focuses on the acoustic aspects of the mother‘s voice, which children may process more efficiently than an unfamiliar voice (e.g. Purhonen, Kilpeläinen-Lees, Valkonen-Korhonen, Karhu & Lehtonen, 2004, 2005). Finally, an exemplar account assumes that initial representations of novel words are context sensitive. Therefore, speaker consistency during both learning and test phase benefits performance on word-learning tasks (e.g. Goldinger, 1998).
Before any of these theories can be tested, we should find clear evidence for the notion that the mother‘s voice indeed facilitates children‘s novel word learning. Therefore, in this study we investigated whether at the group level 24-month-olds are better able to make new word-object associations when they are taught by their mother compared to when they are taught by an unfamiliar female person. Based on the previously found effects of the mother‘s voice during language processing tasks, we expected that it would indeed have a beneficial effect on children‘s novel word-learning abilities. This was tested in an eye-tracking task, with a paradigm closely resembling that of Ma et al. (2011), which looked into the effects of child-directed speech on word-object mapping.
Since previous studies have shown that the course of language development is highly variable between children, we expected to find high variability in the performance on the word-object association task. To explain some of this variation at the individual level, we included two factors that may be related to word-learning abilities.
As a linguistic factor, we looked at the current active vocabulary size, which was indicated by the parents through a questionnaire. Around 18 months of age, children‘s active vocabulary increases very rapidly. This phenomenon is often referred to as the vocabulary spurt (Goldfield & Reznick, 1990) and is thought to be the result of a change in cognitive and linguistic abilities. Children at this age start to make use of several constraints which help them form initial hypotheses about the meaning of a novel word (Markman, Wasow & Hansen, 2003). For instance, the whole-object assumption steers the child towards the hypothesis that a novel label refers to the entire object, rather than a single part or characteristic. Additionally, the mutual exclusivity bias also facilitates word-learning by helping the child realise that each object only has one label (Markman, 1991). Such heuristics make word-learning much more efficient, which may contribute greatly to the word spurt. In fact, there is evidence that the use of these constraints precedes the onset of the word spurt (e.g. Littschwager & Markman, 1994; Markman, 1994). However, the occurrence of the vocabulary spurt is highly variable, as was shown when children‘s vocabulary size was tested at 2-month intervals from 14 to 22 months old (Reznick & Goldfield, 1992). The age of onset of the vocabulary spurt varied from the 14 16 months interval to the 20 22 months interval, and there was even a complete absence of the spurt for five children. We therefore predicted high variability in the active vocabulary scores in our study, and expected these scores to be positively related to the novel word-learning abilities.
As has been discussed above, social non-linguistic abilities, such as joint attention, are positively related to language development. We therefore included a crucial social ability as a non-linguistic factor: the ability to follow eye gaze. This ability is reliably found to arise at 10 to 12 months for gaze shifts that include a head turn (e.g., Scaife & Bruner, 1975). At around 18 months, children are able to follow a gaze shift made by the eyes alone (Corkum & Moore, 1995). However, by using a computerised setting rather than a naturalistic one, gaze following
for eye shifts alone was also demonstrated in infants as young as 3 months old (Hood, Willen & Driver, 1998). At the group level, we looked into children‘s gaze following behaviour for both their mother and an unfamiliar female person. As was discussed above, one of the theories on why children might learn words more easily from their mother is that they may be more sensitive to the social cues provided by their mother compared to an unfamiliar person (social account). If such an account holds true, then this effect may also be found for the social cue of gaze direction. Therefore, we expected children to perform better at the gaze following task when they were looking at their mother‘s face, compared to an unfamiliar face. To date, there are very few studies directly comparing infant gaze following behaviour in response to their mother versus a stranger. In adults, gaze following has been found to be stronger for familiar faces compared to unfamiliar faces, but only for female subjects (Deaner, Shepherd & Platt, 2007). A similar effect has been demonstrated in an ERP study with 4-month-old infants (Hoehl, Wahl, Michel & Striano, 2012). The authors found that a gaze cue towards an object from the infant‘s caregiver reduced the amount of encoding required to visually process that object. No such effect was found for the gaze cue of a stranger. Finally, one eye-tracking study found a differential developmental trajectory of gaze following, such that gaze following performance increased dramatically between 4 to 6 months for interactions with strangers (Gredebäck, Fikke & Melinder, 2010). This increase in performance was delayed for interactions with their mothers by 2 months, after which gaze following performance was better for interactions with the mother compared to a stranger. There is no such data available for children around 24 months of age.
Gaze following abilities have often been linked to language development. For instance, gaze following behaviour at 10 to 11 months significantly predicts vocabulary growth through the second year (Brooks & Meltzoff, 2008), and is positively correlated to vocabulary size at 18 months (Brooks & Meltzoff, 2005). On account of such results, gaze
following abilities are thought to facilitate word-learning, since they allow children to determine what object a speaker is referring to when they are naming it. Therefore, in this study, we directly tested at the individual level the relationship between gaze following performance and word-learning skill in a computerised eye-tracking task. We expected the gaze following performance to be positively related to the performance on the novel word-learning task.
To sum up, this study addresses several questions. At the group level, we ask whether children demonstrate better word-learning skills when they learn new word-objects associations from their mother compared to an unfamiliar person. Secondly, we investigate whether children are more sensitive to social cues from their mother versus an unfamiliar person by looking at their gaze-following behaviour. At the individual level, we ask whether the variation in word-object mapping skill is related to current productive vocabulary size and social non-linguistic gaze following performance.
Method
Participants
Eighty-one monolingual Dutch-learning two-year-olds (44 girls) participated in the study (M age: 24 months 12 days, range 23 months – 25 months 15 days). They were all born after minimally 37 weeks gestation and there were no known language deficits in the immediate family. For the novel word-learning task, the participants were randomly assigned to either the familiar voice condition (M age: 24 months 12 days; range: 23 months 25 months 15 days; 19 boys and 22 girls) or the unfamiliar voice condition (M age: 24 months 12 days; range: 23 moths 3 days 25 months 11 days; 18 boys and 22 girls).
Parents with children in the appropriate age range were approached via a written recruitment letter. Their addresses were provided by the local governments of nearby towns.
They were asked to indicate their interest in the study by emailing or sending a response card. Upon doing so, they were contacted by telephone to make an appointment. The study was approved by the local ethical committee, and parents signed informed consent before the start of the experiment.
Apparatus and Materials
The children sat in car seat mounted on top of a table in front of a 24 inch monitor, which had a resolution of 1920x1080 pixels and a refresh rate of 60 Hz. The centre of the monitor was positioned at the child‘s eye level at a distance of 65 cm. The child‘s mother and experiment leader were sitting on either side of the child, also facing the monitor (positions counterbalanced across subjects). Two loudspeakers were placed left and right behind the monitor and were used to present the sounds in the gaze following task.
Stimuli were presented with MATLAB R2013a and the PsychToolbox tool (version 3.0.11; Brainard, 1997). A Tobii TX300 eye tracker with a 300 Hz sampling rate was mounted below the monitor and was used to collect the gaze positions during the tasks.
Children‘s vocabulary was measured with the ‗Words‘ part of the N-CDI Words and Sentences questionnaire (Zink & Lejaegere, 2003), filled out by the parents. This questionnaire consisted of 112 word items, containing nouns, verbs, propositions and short phrases (e.g., ―bye bye‖, ―good night‖, ―thank you‖). Parents were asked to indicate for each word whether their child used it in their active vocabulary or merely in their comprehensive vocabulary.
Stimuli
Novel word-learning task. For the novel word-learning task, the visual stimuli
Familiarisation trial image Animation screen caption Test trial image
first presented to the child in an animation. The animations (made similar to the ones used in Ma et al., 2011) consisted of 721 frames, which were created in Adobe Photoshop CC 2014. During the animation, the object was presented in a smaller window with a white background at the centre of the screen (size: 720x480 pixels). The object first dropped into the window, moved forwards and backwards, turned 360 degrees, jumped to the right, turned around, jumped to the left and out of sight, and then reappeared, dropping in from the left corner of the window. The familiar objects used for familiarisation with the task and the novel objects during test trials were presented side-by-side as static images, each in a 720x480 pixel window with a white background. The windows were separated by 280 pixels of background. For all trials, the background of the screen was grey. Figure 1 shows the familiar objects and novel objects used in this task.
Figure 1. Examples of different visual stimuli used in the novel word-learning task.
The auditory stimuli in the novel word-learning task were read-aloud sentences that either introduced the novel objects in the training trials or directed the child to look at either one of the objects in the test trials. The sentences were presented in text on the computer screen below the visual stimuli, such that the mother or the experiment leader could read them out loud while still looking at the screen. The speaker was instructed to read the sentences clearly and slowly in child-directed speech. The non-existing Dutch pseudo words ‗gemer‘
Direct gaze Averted gaze left Averted gaze right
and ‗miekel‘ were chosen as the names for the novel objects. They resemble possible Dutch object names both in phonology and morphology and are clearly distinguishable. During their first appearance, the objects were referred to with the indefinite article ―een‖, and in further appearances with the definite article ―de‖.
Gaze following task. For the gaze following task, the visual stimuli consisted of
pictures of faces and target images. Prior to the experiment, three close-up photographs were taken of each mother‘s face, in which only the gaze direction varied (straight ahead, left, and right). These photographs were then edited using Adobe Photoshop CC 2014 in such a way that all background information was taken out. The cut-out image of the face was placed on a grey background. Target images consisted of small, square, schematic and colourful images of objects or animals (e.g. a flower, cow, or circle; 150x150 pixels). Figure 2 shows an example of a direct-gaze stimulus with the two corresponding averted-gaze stimuli. The auditory stimuli were random, short, non-linguistic sounds that accompanied the target image after 1 second, to direct the child‘s attention to the target.
Figure 2. Example of face stimuli used in the gaze following task. Each trial consisted of one direct-gaze image, followed by one of the two averted-gaze images.
Design
The novel word-learning task had one between-subjects condition, namely the familiarity of the voice (mother‘s voice or unfamiliar voice). The name assignments to the
novel objects, the order in which the novel objects were introduced, and the position of the novel objects during the test phase were counterbalanced across participants.
The gaze following task had 2x2 within-subjects conditions, which were the congruency of the gaze cue with the target position (congruent or incongruent) and the familiarity of the face (mother‘s face or unfamiliar face). For the unfamiliar face, pictures of the previous participant‘s mother were used.
Procedure
Novel word-learning task. The novel word-learning task consisted of five phases.
Before the start of each trial, a fixation image of a small star was presented at the centre of the monitor. The experimenter started the trial when the child was looking at the fixation image. During each trial, the speech stimulus that had to be read out by the mother or the experimenter was presented below the visual stimulus on the screen.
During the familiarisation phase, the children were presented twice with an image of two familiar objects side-by-side (apple and book). They were directed to look at one object in the first trial and the other object in the second trial, by having their mother or the experiment leader read out the familiarisation sentence. The order of these familiarisation trials was counterbalanced between participants.
During the training phase, the novel objects were introduced in a short animation, while either the child‘s mother or the experimenter read the training sentences, introducing the object‘s name. Each object was introduced twice, in alternating order, at the centre of the screen, accompanied by their training sentence.
The test phase consisted of two blocks of two trials, separated by the reminder phase. In the test blocks, two static images of the novel objects were presented side-by-side, and the child‘s mother or experiment leader read out the test sentences to direct the child to look at
either one of the objects. Each object occurred twice as target, in alternating order. The position of the objects and order of the target words was counterbalanced across participants. During the reminder phase, the objects were once again presented in a static image at the centre of the screen in alternating order, companied by a short reminder sentence introducing their name again. The task lasted seven minutes in total. Figure 3 shows the visual stimuli and sentences used for each phase of the novel word learning task.
Visual Audio (English) Audio (Original Dutch ) Familiarization
(2 trials: one for each word)
“Apple! Look at the apple! Do you
see the apple? There’s the apple!” “Appel! Kijk naar de appel! Zie je de appel? Daar is de appel!”
“Book! Look at the book! Do you see the book? There’s the book!”
“Boek! Kijk naar het boek! Zie je het boek? Daar is het boek!”
Training
Animations of objects (4 trials; the 2 trials repeat)
“Look! It’s a Gemer! See the Gemer! It’s the Gemer. What is the Gemer doing? The Gemer is going over there. Where is the Gemer going? Where is the Gemer? Gemer! There’s the Gemer!”
“Kijk eens! Het is een Gemer! Kijk! De Gemer! Het is de Gemer. Zie je wat de Gemer doet? Nu gaat de Gemer daar heen. Wat gaat de Gemer doen? Waar is nou de Gemer? Gemer! Daar is de Gemer!” “Look! It’s a Miekel! See the Miekel!
It’s the Miekel. What is the Miekel doing? The Miekel is going over there. Where is the Miekel going? Where is the Miekel? Miekel! There’s the Miekel!”
“Kijk eens! Het is een Miekel! Kijk! De Miekel! Het is een Miekel. Zie je wat de Miekel doet? Nu gaat de Miekel daar heen. Wat gaat de Miekel doen? Waar is nou de Miekel? Miekel! Daar is de Miekel!”
Test block 1
(2 trials: one for each word)
“Gemer! Look at the Gemer! See the Gemer? There’s the Gemer!”
“Gemer! Kijk naar de Gemer! Zie je de Gemer? Daar is de Gemer!”
“Miekel! Look at the Miekel! See the Miekel? There’s the Miekel!”
“Miekel! Kijk naar de Miekel! Zie je de Miekel? Daar is de Miekel!”
Reminder
(2 trials: one for each word)
“Gemer! It’s the Gemer! Look! It’s the Gemer.”
“Gemer! Het is de Gemer! Kijk! De Gemer! Het is de Gemer.”
“Miekel! It’s the Miekel! Look! It’s the Miekel.”
“Miekel! Het is de Miekel! Kijk! De Miekel! Het is de Miekel.”
Test block 2
(2 trials: one for each word)
“Gemer! Look at the Gemer! See the Gemer? There’s the Gemer!”
“Gemer! Kijk naar de Gemer! Zie je de Gemer? Daar is de Gemer!”
“Miekel! Look at the Miekel! See the Miekel? There’s the Miekel!”
“Miekel! Kijk naar de Miekel! Zie je de Miekel? Daar is de Miekel!”
300 500ms 300 500ms 1000ms 1000ms
Gaze following task. In the gaze following task, each trial was started by the
experimenter when the child was looking at the fixation image of a small star. In this case, the fixation image was presented at the horizontal centre of the monitor and at the vertical position of the eyes in the face stimuli (440 pixels from the top of the screen). For each trial, either the familiar face or the unfamiliar face was presented and the gaze cue was either congruent or incongruent with the subsequent target location. Five blocks of 16 trials were presented. Each of the four conditions was presented four times per block, with the target appearing twice on either side of the screen. The order of presentation of the different trials was randomised for each block. The task lasted 10 minutes in total.
Each trial started with the direct-gaze picture of the chosen face (300 500ms), after which the averted-gaze picture was presented at the same position, making it look like the person had shifted her gaze (300 500ms). Then, the face disappeared and a target image was presented at either the left or right side of the monitor (at 75 pixels from the horizontal edge of the monitor). After 1000ms, the target made a clockwise rotation movement, accompanied by a short sound to attract the child‘s attention (1000ms). Figure 4 shows the progression of a congruent trial in the gaze following task.
Data Analysis
The dependent variable for the novel word learning task was the proportion looking time at the target and non-target in each test trial. The gaze following performance in the gaze following task was measured by the difference in reaction times between congruent and incongruent trials.
Fixations were identified using the Identification by 2-Means Clustering algorithm (I-2MC; Hessels, Niehorster, Kemner, & Hooge, in prep.), which is based on a procedure called k-means clustering (where k = 2). It determines whether one or two fixations are present in a small moving window (200ms). Using this moving window makes the algorithm robust to variations in local noise, making it especially appropriate for infant data. Fixations that were separated by less than 0.7 or less than 40ms were merged.
After applying the algorithm, fixations shorter than 100ms were removed for both tasks. In the novel word-learning task, we analysed fixations in the window of 200 2200ms after target word onset to test whether children learnt the novel words-object associations. Test trials were only included if there was at least one fixation on both the target and distracter object after target word onset. Participants were included when at least two test trials met these criteria. In the gaze following task, trials were included when there was at least one fixation on the averted-gaze face (between 540 1380 pixels horizontally and 60 1020 pixels vertically) and the target image. For each participant, trials with reaction times deviating more than two standard deviations from that participant‘s mean reaction time were removed. Only participants with at least two valid trials per condition were included.
Results
NCDI Vocabulary Questionnaire
Data on the current vocabulary of 81 participants was collected using a parent-report questionnaire containing questions on the child‘s active and passive vocabulary (words and sentences they produce or understand, respectively). The average active vocabulary consisted of 64 words (minimum: 11, maximum: 110). This is just above the 50th percentile compared to the norm group (norm group: boys 59 words, girls 63 words; our study: boys 62 words, girls 66 words). No significant difference was observed between boys (61 words, n = 37) and girls (67 words, n = 44), nor between the mother‘s voice condition (65 words, n = 40) and the unfamiliar voice condition (63 words, n = 41) of the novel word-learning task. The active vocabulary correlated nearly significantly with age (r (79) = 0.20, p = .067).
Novel Word-Learning Task
There were 79 participants who contributed enough data points to test whether they learnt the novel word-object associations better from their mother or an unfamiliar person (43 females, 36 males; 40 in mother‘s voice condition, 39 in unfamiliar voice condition; M age: 24 months 13 days). Two participants were excluded due to insufficient data. We calculated the proportion looking times at the target object and the distracter object in the time window of 200 2200ms after target word onset. The performance on the task was then determined by comparing the proportion looking times for both of these object-types. A significant difference between target and distracter looking times indicated a correct word-object mapping, hereafter referred to as the naming effect.
We performed a repeated measures ANOVA with object-type (target and distracter) and block (test bock 1 and test block 2) as within-subjects factors and voice condition (mother‘s voice or unfamiliar voice) as a between-subjects factor to determine word-object
Mother's voice Stranger's voice 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9
Block 1 Block 2 Block 1 Block 2
Pro po rti o n lo o king ti m e Target Distracter
mapping performance in the two voice conditions and to see whether performance increased or decreased as the experiment went on.
Effect of object-type and voice condition. The repeated measures ANOVA showed a
significant naming effect (F (1, 75) = 13.920, p < .001, 2 = 0.16), with a higher proportion looking time for the target object (M = 0.54, SD = 0.14) than for the distracter object (M = 0.42, SD = 0.15). There was no significant main effect of block (F (1, 75) = 2.626, p = .109). The between-subjects effect of voice condition was not significant (F (1, 75) = 1.318, p = .255). However, there was a significant interaction between voice condition, object-type and block (F (1, 75) = 4.208, p = .044, 2 = 0.05).
Follow-up paired-samples t-tests revealed that a significant naming effect is found in both test blocks for the mother‘s voice condition (block 1: t (39) = 2.931, p = .006, 2 = 0.18; block 2: t (39) = 2.460, p = .018, 2 = 0.13). For the unfamiliar voice condition, there was only a significant naming effect in block 2 (t (36) = 2.884, p = .007, 2 = 0.19), whereas no significant effect was found in block 1 (t (36) = -0.596, p = .555). These results are shown in Figure 5.
Figure 5. Proportion looking-times at target and distracter for both voice conditions in block 1 and 2. Error bars reflect one standard error from the mean.
Effect of vocabulary size and age. To check the robustness of the effect of
voice-condition, we tested weather the results would hold when controlling for differences in vocabulary size and age. We therefore performed the same repeated measures ANOVA with the active vocabulary score and age as covariates. Upon doing so, the significant naming effect found in the previous analysis disappeared (F (1, 73) < 1, p = .327). There was no significant interaction of age with the object-type (F (1, 73) < 1, p = .347). On the other hand, the interaction of active vocabulary with object-type neared significance (F (1, 73) = 3.706, p = .058, 2 = 0.05). The three-way interaction of voice-condition, object-type and block remained significant and increased in strength (F (1, 73) = 5.037, p = .028, 2 = 0.07).
Since the active vocabulary score interacted almost significantly with the naming-effect, we performed follow-up analyses for which we split the sample into a low vocabulary and high vocabulary group. In the low vocabulary group (18 in the mother‘s voice condition, 20 in the unfamiliar voice condition), children had an average productive vocabulary of 43 words, which is around the 25th percentile. Children in the high vocabulary group (22 in the mother‘s voice condition, 17 in the unfamiliar voice condition) had an average productive vocabulary of 85 words, which is close to the 75th percentile. The groups differed significantly in age (F (1, 75) = 7.433, p = .008, 2 = 0.09), with children in the low vocabulary group being slightly younger (M age: 24 months 6 days) than children in the high vocabulary group (M age: 24 months 18 days). Paired-samples t-tests showed no significant naming effect for children with low vocabularies, regardless of the voice condition they were in (mother‘s voice: t (17) = 0.713, p = .485; unfamiliar voice: t (19) = 0.485, p = .633). For children in the high vocabulary group, there was a significant naming-effect in the mother‘s voice condition (t (21) = 4.714, p < .001, 2 = 0.51), with a larger proportion looking time at the target (M = 0.61, SD = 0.13) than at the distracter (M = 0.36, SD = 0.15). This was not the case for the unfamiliar voice condition (t (16) = 1.777, p = .095). These results can be found in Figure 6.
Low vocabulary High vocabulary 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
Mother Stranger Mother Stranger
Pro po rti o n lo o king ti m e Target Distracter
Figure 6. Proportion looking-times at target and distracter for children with low and high vocabularies in both voice-conditions. Error bars reflect one standard error from the mean.
Gaze Following Task
The gaze following performance of 64 participants (36 females; M age: 24 months 12 days; 31 children had participated in the mother‘s voice condition; 33 in the unfamiliar voice condition) was determined by looking at the effect of the congruency between gaze direction and target location on the latency between target onset and the first fixation on the target. From now on, this latency is referred to as the reaction time. If the children followed the gaze direction well, they should consequently have faster reaction times on congruent trials compared to incongruent trials. Therefore, subtracting the mean reaction time for congruent trials from the mean reaction time for incongruent trials results in a measure of the congruency-effect, indicating gaze following performance. Seventeen additional participants were excluded from the analysis due to insufficient data.
We performed a repeated measures ANOVA1 with cue-congruence (congruent vs. incongruent) and face-type (mother vs. stranger) as within-subjects factors to determine the presence of a significant congruency-effect as well as a possible difference in gaze following performance for different faces (mother or stranger). Since gender differences are often observed in infant social cognition tasks that involve looking behaviour (e.g., Connellan, Baron-Cohen, Wheelwright, Batki & Ahluwalia, 2000; Lutchmaya & Baron-Cohen, 2002), gender was added as a between subjects factor. Additionally, to check whether the novel word-learning task affected the performance on the subsequently performed gaze following task, we added word-learning condition (mother‘s voice vs. unfamiliar voice) as an additional between subjects factor.
Effect of cue-congruence and face-type. The repeated measures ANOVA did not
show a significant main-effect of cue-congruence (M = 402.1ms for congruent trials and M = 407.6ms for incongruent trials), F (1, 60) = 3.135, p = .082. There was no main-effect of face type (F (1, 60) < 1, p = .674) and no interaction between cue-congruence and face-type (F (1, 60) = 1.634, p = .206).
Effects of gender. The between-subjects main-effect of gender was not significant (F
(1, 60) = 1.141, p = .290). However, cue-congruence did interact significantly with gender, F (1, 60) = 7.874, p = .007, 2 = 0.12. Follow-up paired samples t-tests showed that girls‘ reaction times were significantly faster during congruent trials (M = 390.5ms) compared to incongruent trials (M = 403.8ms), t (35) = -2.759, p = .009, 2 = 0.18. Boys‘ reaction times did not differ significantly between congruent and incongruent trials (t (27) < 1, p = .414).
1 Reaction time data collapsed over all test blocks were used. We did test the effects of attention and learning by
adding block as another within-subjects factor into the repeated measures ANOVA. We did find a significant main effect for block with faster overall reaction times during block 3 and 4 (M = 392.4) compared to block 1 and 2 (M = 422.62), F (1, 59) = 25.257, p < .001, 2 = 0.31. However, the block factor did not interact with any other factors in the analysis. Most importantly, there were no significant differences in congruence-effects between blocks. This indicates that gaze following performance did neither improve nor decline during the course of the task.
Thus, a significant congruence-effect is only observed for girls. These results are shown in Figure 7.
The interaction-effect of face-type and gender was nearing significance (F (1, 60) = 3.232, p = .077). However, follow-up paired-samples t-tests showed that these results could be deemed negligible and should not be considered during the calculation of gaze following performance (see Appendix for details of these analyses).
Figure 7. Mean reaction times (ms) for congruent and incongruent trials for boys and girls. Error bars reflect one standard error from the mean.
Effects of voice condition. Since the novel word-learning task always preceded the
gaze following task, we were curious to see if the assignment to a certain condition in the former task influenced the performance in the latter. Thirty participants had learnt novel words from their mother, while 34 participants learnt words from an unfamiliar experimenter. There was no significant main effect of word-learning condition (F (1, 60) < 1, p = .685). However, it did significantly interact with cue-congruence, F (1, 60) = 4.556, p = .037,
300 320 340 360 380 400 420 440 460 Girls Boys M e an R T i n m s Congruent Incongruent
300 320 340 360 380 400 420 440 460
Mother's voice Stranger's voice
m ea n R T in m s Congruent Incongruent 2
= 0.07. We therefore decided to look into the congruence-effect for both word-learning groups separately. Follow-up paired-samples t-tests indicated that children who had learnt novel words from their mother did not show a congruence-effect during the gaze following task (t (29) < 1, p = .826). On the other hand, children who had learnt novel words from an unfamiliar person had significantly faster reaction times on congruent gaze following trials (M = 400.5ms) compared to incongruent trials (M = 412.9ms), t (33) = -2.560, p = .015, 2 = 0.17. This shows that the word-learning condition influenced the performance on the gaze following task in such a way that a significant congruency-effect was only observed for children who had learnt novel words from an unfamiliar person. These results are shown in Figure 8.
A more complicated role of the word-learning condition was also found in terms of a four-way interaction between word-learning condition, gender, face-type and cue-congruence, F (1, 60) = 5.366, p = .024, 2 = 0.08. However, post-hoc tests revealed that the effects of face-familiarity are again negligible. Therefore we concentrate in the discussion on the two-way interactions between cue-congruence with gender and with word-learning condition, respectively. The results from the post-hoc tests can be found in the Appendix.
Figure 8. Mean reaction times (ms) for congruent and incongruent trials for different novel word-learning conditions. Standard errors are plotted as error bars.
Gaze following performance. In light of the discussed results, we found no reason to
calculate separate gaze following performance scores for different face-types or blocks. We did, however, decide to calculate different scores based on the two different word-learning conditions and genders, since both factors showed to have significant effects on gaze following performance. Subtracting the reaction times for congruent trials from the reaction times for incongruent trials led to a mean gaze following score of -0.89ms for children who had been in the mother‘s voice word-learning condition (n = 30, SD = 21.9, range = -42.95 49.89ms). Children in the unfamiliar voice word-learning condition had a mean gaze following score of 12.37ms (n = 34, SD = 28.2, range -55.77 77.78ms). Boys had a mean gaze following score of -2.97ms (n = 28, SD = 18.9, range -40.70 45.55ms), while girls had a mean score of 13.25ms (n = 36, SD = 28.8 range -55.77 77.78ms).
Relationship Vocabulary Size, Gaze Following Performance, and Word-Learning
We predicted that there would be a large variation in word-object mapping abilities. We furthermore hypothesised that word-object mapping performance would be related to the current active vocabulary size and the performance on the gaze following task. The performance on the word-object mapping task was quantified by the difference score between proportion looking times at the target and distracter. We looked into the correlations between this score with both the active vocabulary score from the NCDI questionnaire and the gaze following scores described above. Since boys and girls performed significantly different on the gaze following task, we looked at correlations for both groups separately. The same goes for the word-learning voice condition groups.
The performance on the word-object mapping task was highly variable. The looking time difference scores ranged from -0.50 to 0.61, and were positively correlated with the
active vocabulary score (r (75) = 0.25, p = .027). The vocabulary scores were showed great variability, with scores ranging from around the 1st percentile (boys 11 and girls 12 words) to above the 99th percentile (boys 100 and girls 101 words).
The gaze following scores were highly variable as well, with scores ranging from -40.70ms to 45.55ms for boys and from -95.35ms to 77.78ms for girls. For children who had been in the mother‘s voice condition, scores ranged from -95.35ms to 49.89ms, and for children in the unfamiliar voice condition from -55.77ms to 77.78ms. The gaze following performance did not correlate with the proportion looking-time difference score for either gender (boys: r (28) = -0.69, p = .717; girls: r (34) = -0.10, p = .526) or either voice condition (mother‘s voice condition: r (29) = -0.13, p = .476; unfamiliar voice condition: r (31) = 0.13, p = .476).
Discussion
This study provided a first look into the possible beneficial effects of the mother‘s voice on 24-month-olds‘ novel word-learning abilities. Specifically, we tested whether children formed new word-object associations more easily when they were taught by their mother compared to an unfamiliar person. We predicted high variability in the word-object mapping performance across individuals, and therefore included two other abilities linguistic and non-linguistic to explain some of this variation. We looked for evidence for a direct relationship between word-learning performance and active vocabulary size. Moreover, we looked into the relationship between the word-object mapping abilities and children‘s ability to follow eye gaze. The latter ability was investigated further at the group level, to test the notion that the mother‘s voice benefits novel word-learning because children are more sensitive to social cues from their mother compared to a stranger.
Novel Word Learning
We found a significant naming effect with a difference of 12% in looking time between the target and distracter image. This indicates that children overall correctly learnt the new word-object mappings used in our study. Of particular interest was the interaction of the learning context (i.e. familiar or unfamiliar voice) with this naming effect. Although we did not find a significant interaction when we looked at the task as a whole, the learning context did interact with the naming effect when we looked at the test blocks separately. Children who had learnt new word-object pairs from their mother showed correct target-distracter differentiation in both test blocks, whereas this effect was only found in the second test block for the group who had learnt the pairs from an unfamiliar person.
These results show that children are able to learn new word-object associations from their mother after being exposed to them for a short period in the training phase. On the other hand, children who learnt the word-object pairs from an unfamiliar person only showed correct recognition of the pairs after the reminder trials, in which each pair was shortly introduced again. Previous studies on rapid word learning show that children learnt new word-object associations from an unfamiliar person after only a few repetitions. This effect has been found in 13 and 18-month-olds after only nine repetitions of each word-object pair (Woodward, Markman & Fitzsimmons, 1994), and in 15-month-olds using six repetitions, with pre-recorded audio (Schafer & Plunkett, 1998). A more recent study found rapid word-object mapping only in 13-month-olds, and not in 17-month-olds (twelve live-speech repetitions).
These findings suggest that the 24-month-olds in our study who heard the word-object pairs from an unfamiliar person should have been able to learn them correctly after our two training trials. Each trial contained nine repetitions of the novel word-object pair, making it
eighteen repetitions of each pair by the end of the training phase. It is unlikely that merely the addition of three extra instances of the word-object pairing in the reminder trials caused the performance of children in this voice condition to increase suddenly. The theories on the beneficial influence of the mother‘s voice on novel word learning as discussed in the introduction provide another possible explanation. The social account emphasises the role of social clues during the word-learning situation. It states that children may be able to understand the intention of the situation better when they are being taught by a familiar person. Extrapolating from this, it is possible that it is not the lack of repetitions causing the children in the unfamiliar voice group to show delayed learning effects, but rather it is the difference in the social nature of the training and reminder trials. The training trials contained an animation paired with a narration, making the situation quite similar to when parents read a picture book to their child or talk along with an animation at home. Since children are more familiar with being in such a situation with their mother than with an unfamiliar person, they may pick up the new word-object associations more easily when the narration is done by their mother compared to a stranger. It is possible that the children in the unfamiliar voice condition therefore needed more time to learn the associations, or that the more straightforward nature of the reminder trials (with a short sentence and a static image) was more effective. Such an explanation is purely speculative and would need further experimentation to be supported, for instance by testing whether we find similar results using pre-recorded rather than live auditory stimuli, thus removing some of the social context.
The fact that children in the mother‘s voice condition did learn the novel word-object pairs more easily is in accordance with previous studies which showed that the mother‘s voice had a beneficial effect on the recognition of familiar words (Bergelson & Swingley, 2012; Parise & Csibra, 2012) and the acquisition of new word forms (without meaning; Barker & Newman, 2004). Our paradigm was based on a study by Ma et al. (2011), which showed that
child-directed speech had a beneficial effect on word-object mapping in 21-month-olds. However, they did not find the same effect in 27-month-olds, as children in this age group learnt new words from adult-directed speech just as well as from child-directed speech. The authors argued that by this age, children have had enough language experience to learn new words without the help of facilitating perceptual cues (Hollich et al., 2000). Since the age of children in the current study falls in between the two age groups in the study by Ma et al., it could be that the beneficial effect of the mother‘s voice is not as clearly pronounced as it would be in younger children. It would therefore be interesting to test whether younger children benefit even more from learning words from their mother instead of an unfamiliar person. Additionally, it would be interesting to see if the effect disappears at an older age, as this would be in concurrence with the theory that older children do not rely on beneficial perceptual cues as much as younger children do.
Although we did find a beneficial role of the mother‘s voice at the group level, this does not mean that children do not learn new word-object associations from unfamiliar people at all. When we looked at the individual word-learning performance, the scores were highly variable, and the naming effect was not limited to the mother‘s voice condition. This large amount of variation was expected, based on the fact that vocabulary development in general shows great differences between individuals (Fenson et al., 1994; Reznick & Goldfield, 1992).
Novel Word Learning and Vocabulary Size
A large amount of variation was also found in the productive vocabulary measurements. The average scores for boys and girls were just above the 50th percentile compared to the norm group, with scores ranging from 11 to 110 words. We did not find any significant sex differences in vocabulary size, whereas previous studies did report higher
scores for girls compared to boys (e.g., Bornstein, Leach & Haynes, 2004; Galsworthy, Dionne, Dale & Plomin, 2000). However, other authors argue that the effect of gender in these language measures is so small that it should be deemed negligible (only about 1 word difference at 18 months; Berglund, Eriksson & Westerlund, 2005). The small, non-significant, advantage for girls in the present study therefore conforms to previously found gender effects on vocabulary size.
The productive vocabulary size correlated significantly with the performance on the word-object association task. The direct match between vocabulary size and word-learning abilities supports the theory that the vocabulary spurt around 18 months is due to certain heuristics that children around this age start to use, such as mutual exclusivity and whole-object assumptions, as was discussed in the introduction. These heuristics make word-learning much more efficient and effortless, resulting in a rapid growth of the child‘s vocabulary (Markman, 1991).
To further investigate the role of vocabulary size in our task, we tested whether the findings from the novel word-learning task would hold after controlling for age and current active vocabulary size. Interestingly, the overall effect of naming disappeared, while the interaction of voice, object-type and block increased in strength. Moreover, the vocabulary size interacted nearly significantly with the proportion looking times. This indicates that the overall naming effect that we initially found was strongly influenced by the children‘s current vocabulary, as was already hinted at by the significant correlation discussed above. Based on these findings, we then ran an analysis in which we split the sample in a low vocabulary and high vocabulary group. The results further confirmed the influence of vocabulary size in this study: a significant naming effect was only found for the group with a high current active vocabulary. Regarding the influence of the mother‘s voice, we found a beneficial effect on word-object mapping only for children that already had a large vocabulary.
Contrary to our results, Ma et al. (2011) did find a significant naming effect in their low vocabulary group (21-month-olds). Whereas children with a large vocabulary performed well in both the child- and adult-directed speech conditions, the low vocabulary group only learnt the novel words when they were taught in child-directed speech. Since our study used child-directed speech in both voice conditions, it is curious that we did not find any evidence of word-learning for our low vocabulary group. Our experimental design differed from the previous study in that we used live speech stimuli instead of pre-recorded sentences. It seems that this more social learning context made the task harder for the children, making it impossible for children with less developed language skills to learn the new word-object associations. Another possibility is that our live child-directed speech sentences did not differ as much from adult-directed speech as in the study by Ma et al., in which the pre-recorded sentences were analysed for the correct child-directed speech characteristics such as a high fundamental frequency before they were used. We believe this explanation is less plausible, since the mothers and experimenters in our study were explicitly instructed to use child-directed speech and were corrected during the familiarisation phase if their reading did not conform to this expectation.
All in all, the added difficulty of our task made it only possible to learn new word-object associations for children with more developed language skills, provided that they could draw on the beneficial effects of the mother‘s voice during the training phase.
Novel Word Learning and Gaze Following Performance
The second factor we included to explain some of the variance in word-learning ability was gaze following behaviour. We included trials containing the mother‘s face or a stranger‘s face in order to assess whether children are more sensitive to social cues from their mother versus those of a stranger. Evidence in favour of this hypothesis would support the theory that
children learn new words more easily from their mother because they are better aware of the social, word-learning, situation.
We found a congruence effect only for girls, independent of the face-type. Boys did not seem to follow eye gaze for either face. The size of the gaze cueing effect that we found for girls is similar to findings in two previous studies with typically developing children around this age. Children in these studies were 26 months (Chawarska, Klin & Volkmar, 2003) and 36 months old (Johnson et al., 2005), and had a difference in reaction times between congruent and incongruent trials of about 12ms in both studies, which is very close to our results of 13.25ms for girls. The superior performance of girls is not surprising given previous studies on gaze following behaviour. In adults, greater cueing effects are found for females compared to males (Bayliss, Pellegrino & Tipper, 2005; Alwall, Johansson & Hansen, 2010). Studies on social cognition in general have often found gender differences in infants as well. For instance, observational studies show that girls initiate more joint attention episodes than boys (Mundy et al., 2007) and make more eye-contact with their parents (Leeb & Rejskind, 2004; Lutchmaya, Baron-Cohen & Raggatt, 2002). Finally, studies on infant looking behaviour showed that girls prefer to look at faces (social objects) over a mechanical object (mobile or car), while the effect for boys was the other way around (Connellan, Baron-Cohen, Wheelwright, Batki & Ahluwalia, 2000; Lutchmaya & Baron-Baron-Cohen, 2002). Our observed gender difference in gaze following performance therefore fits in well with these previously demonstrated differences in adults and infants.
The literature on gaze following behaviour for familiar and unfamiliar faces is less abundant, although gaze cueing has been shown to be stronger for familiar versus unfamiliar faces in both adults (only for females; Deaner, Shepherd & Platt, 2007) and infants (4-month-olds; Hoehl, Wahl, Michel & Striano, 2012). It is therefore surprising that we did not find an effect of face familiarity in our gaze following task. However, Hoehl et al. used an entirely
different measurement of gaze following (ERP components) and tested children that were 20 months younger that the children in the current study. It is therefore impossible to directly compare these results with ours. Unfortunately, there are no studies on the effect of face familiarity on gaze cueing including children closer to 24 months old. A possible cause for the absence of a face familiarity effect in our study is the nature of our visual stimuli. Gaze following studies commonly but not exclusively use visual stimuli in which a lot of information about the person is visible, either by using a live setting or by including a person‘s hair and shoulders in the pictures that are used. On the other hand, our study used the cut-out of a face only. This may have diminished the effects face familiarity, either because the smaller amount of information present in our stimuli made differentiating between the mother and an unfamiliar female more difficult, or because our stimuli did not reflect a naturalistic setting.
Although we speculated above that the differences in word-learning performance between the mother and unfamiliar voice groups may be caused by the social setting of the training trials, the absence of any effects of face familiarity in the gaze following task does not support the notion that children are more sensitive to social cues from their mother compared to a stranger. It is possible that the gaze following task did not tap into the same social cues that were at work during the novel word-learning task. On the other hand, it is possible that the beneficial effect of the mother‘s voice does not depend on social factors at all. Therefore, we once again suggest repeating the word-learning task with pre-recorded speech, to see whether the effects of the mother‘s voice are also found without the social aspects of the task.
At the individual level, we found a large amount of variation in the gaze following performance. We expected this variation to be related to the performance on the novel word-learning task, based on previous studies which showed a link between gaze following abilities
and vocabulary size (Brooks & Meltzoff, 2005; Brooks & Meltzoff, 2008). However, we did not find a correlation between the performance on the gaze following task and the novel word-learning task, even when we looked at the performance of boys and girls separately. This could be the result of the paradigm used to measure word-learning abilities. Because the novel objects were presented on the computer screen where the child was already looking, the children were not required to shift their attention to the novel object based on the adult‘s eye gaze to be able to perform well on the novel word-learning task. This may have made it less likely that we would find a relationship between the performance on the word-learning and gaze following tasks. It would be interesting to test whether such a link can be demonstrated in a more naturalistic word-learning paradigm where such shifts of attention are required.
Although we did not find any correlations between gaze following and word-learning skill, we did see that the performance on the gaze following task was influenced by the condition in which the children had learnt new words in the preceding novel word-learning task. A significant congruence-effect was only found for children who had just learnt novel word-object associations from an unfamiliar person. Children who had learnt words from their mother did not seem to follow the eye gaze in the gaze following task. This shows that the social context in which the children learnt the novel words influenced the performance on a subsequent non-learning social task. Perhaps the extra (social) effort required by children in the unfamiliar voice condition made them more prepared for the social task ahead. These results should be taken into account in future studies that include multiple tasks.
Conclusion
Altogether, this study provides the first direct evidence that the mother‘s voice has a facilitating effect on novel word learning, albeit only for children with a large current productive vocabulary. This opens doors to further research into the underlying cause of this
effect, which may lead to further understanding and prevention of language delays. Our gaze following data did not provide any evidence for the notion that this beneficial effect could be caused by children‘s heightened sensitivity to social cues from their mother compared to a stranger. Repeating this study with pre-recorded sentences, thus eliminating the social cues, could give more insight in the plausibility of this social account.
At the individual level, we demonstrate a direct positive relationship between word-learning abilities and children‘s current active vocabulary size, supporting the notion that the large increase in vocabulary size around the second year is due to the development of more efficient word-learning strategies.
Appendix
Follow-up analyses in gaze following task
Effect of face-type. The repeated-measures ANOVA showed an interaction-effect
between face-type and gender that neared significance, F (1, 60) = 3.232, p = 0.77, 2 = 0.05. Post-hoc paired-samples t-tests revealed that this interaction was due to the opposite effects of face-type for both groups, rather than to significantly large differences in reaction times between face-types. In fact, boys have faster reaction times when looking at their mother‘s face compared to a stranger (M = 410.7ms and 415.8ms respectively), whereas girls are faster on trials containing a stranger‘s face compared to their mother‘s face (M = 393.4ms and 401.3ms respectively), but neither effects turn out to be significant (t (27) = -1.390, p = .176 for boys and t (35) = 1. 623, p = .113 for girls).
Interaction of word-learning condition, gender, face-type and cue-congruence.
The repeated-measures ANOVA showed a four-way interaction between word-learning condition, gender, face-type and cue-congruence, F (1, 60) = 5.366, p = .024, 2 = 0.08. These results were further analysed by performing paired-samples t-tests for 2x2 between-subjects groups (word-learning condition x gender), looking at reaction time differences between congruent and incongruent trials for both face-types (mother and stranger).
The analyses showed very similar congruency-effects for both face-types in each group. After Bonferroni correction, boys who had learnt words in either voice condition did not show significant congruency-effects for trials containing either face-type. Girls who had learnt words from their mother also did not show any congruency-effects for either face. Girls who had learnt words from an unfamiliar person did show a significant congruency-effect for both faces, although the effect was much more pronounced for their mother‘s face. These results are summarised in Table 1. Note that the p-value of .046 for boys in the mother‘s voice
* condition on stranger‘s face trials is not an indication for gaze following, since their reaction times are faster for incongruent trials than for congruent trials.
These results indicate that the effect of face-type in the four-way interaction is very small and is therefore deemed negligible. This is in concurrence with the discussed interaction between face-type and gender (under effects of gender in the results section), which was also deemed negligible. The post-hoc tests in table 1 also show that the only significant congruency-effect is observed for girls, which has already been discussed in the results section under effects of gender. Furthermore, it clearly shows that the congruence-effect is only observed for the unfamiliar voice word-learning condition, which has already been discussed under effects of voice condition in the results section. Altogether, the four-way interaction does not offer any relevant or new results concerning gaze following performance. Table 1.
Results from Paired-Samples T-Test of Cue-Congruence (RT Incongruent vs. RT Congruent) for Different Face-Types and for Four Between-Subjects Groups (Word-Learning Condition x Gender)
References
Alwall, N., Johansson, D., & Hansen, S. (2010). The gender difference in gaze-cueing: Associations with empathizing and systemizing. Personality and Individual Differences, 49(7), 729-732.
Barker, B. A., & Newman, R. S. (2004). Listen to your mother! The role of talker familiarity in infant streaming. Cognition, 94(2), B45-B53.
Bayliss, A. P., Pellegrino, G. D., & Tipper, S. P. (2005). Sex differences in eye gaze and symbolic cueing of attention. The Quarterly Journal of Experimental Psychology, 58(4), 631-650.)
Beauchemin, M., González-Frankenberger, B., Tremblay, J., Vannasing, P., MartínezMontes, E., Belin, P., ... & Lassonde, M. (2010). Mother and stranger: an electrophysiological study of voice processing in newborns. Cerebral Cortex, 21(8), 1705-1711.
Bergelson, E., & Swingley, D. (2012). At 6–9 months, human infants know the meanings of many common nouns. Proceedings of the National Academy of Sciences of the United States of America, 109, 3253–3258.
Berglund E, Eriksson M & Westerlund M (2005). Development and Aging: Communicative skills in relation to gender, birth order, childcare and socioeconomic status in 18-month-old children. Scandinavian Journal of Psychology, 46, 485-491.
Bornstein M H, Leach D B & Haynes O M (2004). Vocabulary competence in first- and secondborn siblings of the same chronological age. Journal of Child Language, 31, 855-873.
Brainard, D. H. (1997). The psychophysics toolbox. Spatial vision, 10, 433-436.
Brooks, R., & Meltzoff, A. N. (2005). The development of gaze following and its relation to language. Developmental science, 8(6), 535-543.
