On a crowded city street, we automatically attribute the sounds of cars to the cars we see driving past, and not to the motorcycles or trucks on the same road. Similarly, we assign the voices we hear to the pedestrians around us, and not to the dogs those pedestrians are walking. As adults, we cope with these everyday challenges effortlessly, but how do infants first learn to match what they see with what they hear?

When young animals are exposed to new stimuli, their brains undergo changes. Similar changes only occur in adult animals if they deliberately pay attention to the stimuli and if they are associated with rewards. Rohlf, Habets et al. therefore predicted that human infants would automatically learn to associate sights and sounds upon being passively exposed to them. Adults, on the other hand, would learn these associations only if explicitly asked to do so.

To test this prediction, Rohlf, Habets et al. presented tones and colored shapes to 6-month-old infants and healthy adult volunteers while using scalp electrodes to monitor the electrical activity in their brains. Certain shapes and tones occurred frequently together, whereas other combinations of the same stimuli were rare. The 6-month-olds consistently outperformed the adults in associating the tones and shapes: the electrical activity in the infant brains reliably distinguished between common versus rare combinations. Adult brains made this distinction only when the adults were asked to pay attention to the tone-shape combinations as part of a task.

This high sensitivity to combinations of sights and sounds that regularly occur together enables infants to quickly learn about the world around them. As adults will have done this previously, the most effective strategy for adults is to update their existing knowledge only when such learning enables them to achieve a goal. Further research is needed to find out what happens in the brain to cause this change in learning strategy. Understanding how learning differs in infants and adults will help identify stages of development in which the brain learns particularly easily. This may ultimately help us optimize learning strategies for individuals of different ages.

After birth infants are immediately exposed to a sensory world comprising input of multiple sensory modalities. The developing brain must adapt to the statistical properties of the sensory environment (Fiser et al., 2010) since genetically defined neural circuits are usually crude. Indeed a high sensitivity of infants to statistical regularities within single sensory systems has often been demonstrated (Fantz, 1964; Saffran et al., 1996; Fiser and Aslin, 2002; Bulf et al., 2011). The seminal study of Saffran et al. (1996) reported that eight-month-old infants quickly learn transitional probabilities between syllables by pure exposure to an artificial language. This ability was interpreted as a basic mechanism allowing infants to segment a language. Similar results were found for non-linguistic auditory sequences and for visual patterns (Saffran et al., 1999; Fiser and Aslin, 2002), demonstrating a modality independent sensitivity of infants to statistical patterns in their sensory environment which moreover is not unique to linguistic material. For example, in the visual domain, there is strong evidence that infants are able to implicitly learn subtle statistical relationships among visual objects (Fiser and Aslin, 2002; Bulf et al., 2011; Kirkham et al., 2002). Nine-month-old infants who were exposed to multi element visual scenes, showed greater interest in element pairs which co-occurred more frequently than in pairs which co-occurred less frequently. Moreover, the infants were sensitive to the predictability between elements of the pairs as manifested by the conditional probability relations between these elements (Fiser and Aslin, 2002). The ability to extract statistical patterns of visual stimuli was found even in younger age groups (Kirkham et al., 2002); two-, five-, and eight-month-old infants were habituated to sequences of discrete visual stimuli whose ordering followed a statistical predictable pattern. Subsequently the infants were shown the previously encountered pattern alternating with a novel pattern of identical stimulus components. Infants of all age groups looked longer at the novel sequences providing evidence for the detection of visual statistical regularities at an early developmental stage. These results suggest that infants own powerful mechanisms for extracting the statistical properties of their sensory input without any instructions, explicit feedback, or intentional awareness (Lany and Saffran, 2013; Krogh et al., 2012).

The ability of infants to detect crossmodal statistical regularities within their sensory environment is less well understood, but some basic multisensory abilities, such as multisensory temporal synchrony detection seem to exist within the first month of life (Lewkowicz, 1992). In the next months the capability to perceive higher-level and more complex multisensory relations starts to develop. For example, at the age of six months infants were shown to perceive duration-based (Lewkowicz, 1992) and spatio-temporal based crossmodal relations (Scheier et al., 2003). Furthermore, there is evidence that similar to adults, infants take advantage of crossmodal events in terms of a better discrimination and a faster responsiveness to bimodal compared to unimodal information (Bahrick et al., 2004; Lewkowicz and Kraebel, 2004). First evidence for multisensory facilitation was found in eight-month-old infants as indicated by faster eye movements to spatially aligned auditory and visual cues compared to eye movements to each of these stimuli alone (Neil et al., 2006). Moreover, other studies revealed multisensory benefits for perceptual learning in infants (Bahrick and Lickliter, 2000; Frank et al., 2009). Five-month-old infants were habituated to either an audio-visual rhythm or the same rhythm presented unimodally. In the crossmodal condition, infants were able to discriminate between the familiar and a novel rhythm, whereas no discrimination was observed for the unimodal stimuli (Bahrick and Lickliter, 2000). Corresponding results were found for the learning of an abstract rule in five-month-old infants: they were able to learn the sequence if defined by redundant visual shapes and speech sounds but not if only one sensory modality was involved (Frank et al., 2009). These results suggest that infants are able to learn and use associations between auditory and visual stimuli. However, it must be taken into account that the multisensory effects in infants were not tested against statistical facilitation (probability summation, see Miller, 1982).

Several studies on crossmodal association learning have reported that infants at the age of three months, but not younger, were able to learn specific voice-face combinations; infants were habituated to different unfamiliar voice-face pairings. In the post-familiarization test the infants showed higher attention to the learned voice-face pairs as compared to the novel combinations. The latter category comprised a voice and a face they had heard and seen previously, but the combination of the voice and face was new (Brookes et al., 2001; Bahrick et al., 2005). More recently, near-infrared spectroscopy (NIRS) and event-related potentials (ERPs) were used to test whether infants are able to learn crossmodal associations between arbitrary auditory and visual stimuli. Emberson et al. (2015) used an audio-visual omission paradigm with six-month-old infants and found similar visual cortex activation as a response to an auditory stimulus alone, which had been previously combined with a visual stimulus, as for the presentation of the same visual stimulus. The authors interpreted their findings as evidence for top-down mechanisms to be in place as early as six month of age. Kouider et al. (2015) exposed twelve-month-old infants to pictures of faces paired with one sound and pictures of flowers paired with a second sound. During the test phase the sound preceded the visual stimulus and was either congruent or incongruent with the learned combinations (additionally no sound was used in one third of the trials). An enhanced early positive ERP for congruent visual stimuli as well as an enhanced late negative ERP for incongruent visual stimuli were found. Both studies demonstrate that infants are able to learn crossmodal combinations to which they were exposed. However, none of these studies used an adult control group. Thus, it remains an open question of whether developmental and adult crossmodal learning recruit similar mechanisms. In this context it is interesting to notice that Janacsek et al. (2012) demonstrated superior implicit statistical learning of visual sequences in young children (<12 years) compared to older children and adults; a follow-up study indicated that this advantage was lost when they became more reliant on explicit learning (Nemeth et al., 2013).

Based on non-human animal studies it has been proposed (Keuroghlian and Knudsen, 2007) that developmental and adult plasticity, and thus learning, differ due to different brain states: during the sensitive phase molecular mechanisms dominate that allow for quick and extensive functional and structural synaptic plasticity (synaptogenesis, synaptic strengthening and elimination) allowing the emergence of a functional adaptive connectivity. By contrast, in adulthood these functionally tuned and to some degree stabilized neural circuits undergo adaptations when relevant to the system. Such age dependent changes from developmental to adult plasticity are impressively demonstrated by a study on auditory cortex plasticity in rats: while passive exposure to sounds of a specific frequency results in a permanent reorganization of auditory cortex during the sensitive phase, adult rats reorganized only those aspects of the auditory cortex which were task relevant: for example, rats were exposed to sounds which varied both in sound frequency and level. When they had to discriminate the sounds with respect to sound frequency the frequency representation of auditory cortex changed while the level representation changed when level rather than sound frequency was task relevant (de Villers-Sidani et al., 2007). These findings suggest that adult learning depend to a larger degree on attention and context such as task relevance and reward expectations (Keuroghlian and Knudsen, 2007; Bavelier et al., 2010). This hypothesis was supported by Riedel and Burton (2006) who investigated whether learning of auditory sequences is influenced by task demands; when using a serial reaction time task related to a feature of the auditory stimulus they found learning effects in adult participants while a passive exposure did not result in learning. Similarly, the statistical relations of concurrently presented visual streams were only learned by adults for the attended but not for the unattended streams (Turk-Browne et al., 2005). Emberson et al. (2011) extended these findings by providing evidence in adults that attention was necessary for implicit statistical learning in both the visual and auditory modality.

In the present study we investigated multisensory associative learning in infants and adults to test the hypothesis that infants as young as six months are not only able to learn arbitrary auditory-visual associations but that their sensitivity to crossmodal statistics is even higher compared to adults when crossmodal associations are passively encountered. Thus, in the first experiment we included a group of six-month-old infants (Experiment 1a) and a group of young adults (Experiment 1b). While recording the electroencephalogram (EEG), we presented two frequently occurring audio-visual standard combinations (A1V1, A2V2, p=0.35 each, ‘Frequent standard stimuli’), two rare recombinations of the ‘Frequent standard stimuli’ (A1V2, A2V1, p=0.10 each, ‘Rare recombined stimuli’) and one rare audio-visual combination of an infrequent auditory and an infrequent visual stimulus (A3V3, p=0.10, ‘Rare deviant stimuli’). Recombining the auditory and visual elements of the ‘Frequent standard stimuli’ controls for the likelihood of the auditory and visual elements of the employed crossmodal stimuli. Thus, in order to detect ‘Rare recombined stimuli’ it is necessary to have learned the precise crossmodal combination. By contrast, the likelihood of both the visual and the auditory elements of ‘Rare deviant stimuli’ were lower than for all other auditory and visual elements. Therefore, the present experimental design allowed us to differentiate between the processing of the likelihood of sensory elements (‘Frequent standard stimuli’ vs. ‘Rare deviant stimuli’) and the processing of the conditional probabilities of crossmodal combinations (‘Frequent standard stimuli’ vs. ‘Rare recombined stimuli’).

We predicted ERP differences between the ‘Frequent standard stimuli’ and ‘Rare deviant stimuli’ in both infants (Cheour et al., 2000) und adults (Schröger and Wolff, 1996; Näätänen and Alho, 1995). In contrast, we hypothesized that only infants display an ERP difference for ‘Frequent standard stimuli’ vs. ‘Rare recombined stimuli’ due to a higher sensitivity to crossmodal statistics during infancy.

The goal of the present study was to test for a higher sensitivity of infants as compared to adults to crossmodal statistics and to compare the mechanisms of crossmodal association learning in infants and adults. We conducted four ERP experiments in which infants and adults were exposed to audio-visual stimulus combinations with different probabilities. We presented ‘Frequent standard stimuli’ (A1V1, A2V2, p=0.35 each), rare recombinations of the ‘Frequent standard stimuli’ (A1V2, A2V1, p=0.10 each, ‘Rare recombined stimuli’), and a rare deviant audio-visual combination with an infrequent auditory and visual element (A3V3, p=0.10, ‘Rare deviant stimuli’). While infants passively learned the crossmodal combinations, adults did not. Adults’ ERPs to ‘Rare recombined stimuli’ and to ‘Frequent recombined stimuli’ differed only when crossmodal combinations were task relevant. In contrast, all groups, irrespectively of learning context, showed a sensitivity to the probability of sensory elements, that is, for ‘Rare deviant stimuli’. Table 2 graphically summarizes the main results of all four experiments.

Table 2

	Early time window		Late time window
	Rare deviant - Standard stimuli	Rare recombined - Standard stimuli	Rare deviant - Standard stimuli	Rare recombined- Standard stimuli
Experiment 1a (Infants)	200 – 420 ms	n.s.	420 – 1000 ms	420 – 1000 ms
Experiment 1b (Adults)	180 – 220 ms	n.s.	250 – 1000 ms	n.s.
Experiment 2a (Adults)	80 – 160 ms	n.s.	250 – 850 ms	n.s.
Experiment 2b (Adults)	80 – 160 ms	n.s.	250 – 850 ms	250 – 850 ms

Studies using artificial languages or visual artificial scenes have repeatedly demonstrated that infants develop a sensitivity to the likelihood of events as well as to conditional probabilities (Krogh et al., 2012; Aslin, 2014), partially as early as at the age of two months (Kirkham et al., 2002). Two recent studies addressing crossmodal statistical learning found that six-month and twelve-month-old infants learned to predict a visual stimulus based on a preceding auditory stimulus (Emberson et al., 2015; Kouider et al., 2015). While Kouider et al. (2015) demonstrated that infants at the age of twelve months were able to learn an association between an arbitrary sound and a visual object category (faces vs. flowers), they did not include an adult control group and were thus not able to demonstrate differences in learning between adults and infants, nor were they able to distinguish processes related to the detection of crossmodal combinations and processes related to the familiarity with certain sensory elements.

Thus, the present study extends previous research by showing that the conditional probabilities of crossmodal combinations were extracted by infants as young as six months after a short exposure period while adults failed to learn crossmodal statistics under this condition. It is important to notice that we controlled for the likelihood of the auditory and visual elements of the employed crossmodal stimuli by infrequently recombining the auditory and visual elements of the ‘Frequent standard stimuli’. We provide ERP evidence demonstrating that the processing of the conditional probabilities of crossmodal combinations and the processing of the likelihood of sensory elements can be dissociated: in infants, ‘Rare recombined stimuli’ elicited a left negative potential starting at about 420 ms post-stimulus while ‘Rare deviant stimuli’ elicited a right lateralized positivity starting at 200 ms post-stimulus (Experiment 1a). Adults tested under identical conditions were only sensitive to ‘Rare deviant stimuli’, which differed from ‘Frequent standard stimuli’ in the frequency of their auditory and visual elements (Experiment 1b, ERP effect starting 180 ms post-stimulus) but not for rare crossmodal stimuli which only differed from the ‘Frequent standard stimuli’ in the way the auditory and visual elements were combined. These results demonstrate that infants were able to learn arbitrary crossmodal associations as early as six months of age and thus much earlier than suggested by the study of Kouider et al. (2015) (see Emberson et al., 2015). This finding is in line with behavioral studies employing natural stimuli, which showed that infants from three months onwards were able to learn specific face-voice-parings (Brookes et al., 2001; Bahrick et al., 2005). Our results extended these behavioral findings by providing first evidence that the learning of crossmodal statistics in infancy is particularly sensitive and superior to adults when crossmodal stimuli are not task relevant. It could be argued that the signal to noise ratio of the ERPs in adults was not sufficient in Experiment 1b to demonstrate crossmodal learning in the adult group. However, such an account is highly unlikely given that an effect of smaller size were detected in Experiment 1b and the fact, that in Experiment 2a an ERP difference between ‘Frequent standard stimuli’ and ‘Rare recombined stimuli’ was not significant either despite a much higher signal to noise ratio in comparison to Experiment 1b.

Thus, our results provide evidence that crossmodal statistics are better implicitly learned in the developing than in the adult system. An enhanced sensitivity for low-level statistical patterns during development compared to adulthood has been reported by other studies as well. For example, Janacsek et al. (2012) and Nemeth et al. (2013) demonstrated that children are superior in implicit statistical learning of sequences compared to adults but later lose this advantage and become more reliant on explicit learning. A similar developmental time course was found in a study of Jost et al., 2011, who had investigated the neurophysiological correlates of visual statistical learning in children and adults: children showed learning related ERP effects earlier in the acquisition phase than the adult group indicating that they had quicker acquired the statistical structurer. It is, however, important to notice that not all studies investigating statistical learning during development have found enhanced learning performance in infants or children as compared to adults. For example, Saffran et al. (1996), 1999) reported similar abilities in eight-month-old infants and adults in the extraction of the underlying statistical structure of auditory sequences. Other studies observed better learning for older children and young adults than in younger age groups (Maybery et al., 1995; Fletcher et al., 2000; Kirkham et al., 2007). Arciuli and Simpson (2011) tested children between the age of 5 and 12 years in a visual triplet learning task and reported better learning with increasing age. At first glance, these findings seem to be at odds with the present results. However, a closer look at the employed paradigms suggests that these different outcomes might be related to the complexity of the implemented statistical patterns. For example, Arciuli and Simpson (2011) tested the incidental learning of four visual base triples which draws to a much larger extend from working memory than learning the conditional probability of two audio-visual combinations as used in the present and previous crossmodal learning studies in infants (Emberson et al., 2015; Kouider et al., 2015). Indeed, it is well known that working memory improves during childhood (Zelazo et al., 2008) and thus working memory demands might have been the limiting factor in some studies (e.g. Arciuli and Simpson, 2011). In addition, triples are usually embedded in a continuous stream while the crossmodal stimuli of the present study were individually presented with relatively long interstimulus intervals, thereby clearly demarking the individual events. Furthermore, studies have revealed that the ability to extract statistical patterns from sensory input during infancy improves from the simple tracking of event probabilities early in the development (from three months onwards, see Fantz, 1964) to the learning of more complex and higher-level statistical patterns at a later developmental stage (from twelve months onwards, see Gómez and Maye, 2005). Thus, what most likely declines during development seems to be the sensitivity to simple conditional probabilities (Janacsek et al., 2012). Janacsek et al. (2012) speculated that a decline in the sensitivity to ‘base probabilities’ is necessary for the acquisition of higher order representations and a switch to model-based (in contrast to model-free) learning.

In line with this suggestion, adults did not learn crossmodal statistics when they were irrelevant for the task but became sensitive to them when a specific crossmodal combination was of behavioral relevance. Studies in non-human animals have suggested that during the sensitive phase, neural networks are elaborated in response to a pure exposure to the environment while during later development and in adulthood learning is context-specific and depends on task relevance (e.g. reward) and instructions (Keuroghlian and Knudsen, 2007). Currently, we can only speculate about the neural underpinnings of the age-dependent neuroplasticity as observed in the present study. As noted by Dehaene-Lambertz and Spelke (2015) feedforward connectivity seems to be to a larger degree genetically determined than feedback connectivity and the latter seems to be mostly experience dependent. The detection of ‘Rare recombined stimuli’ was associated with a relatively late ERP effect in both infants and adults. Indeed, multisensory binding has been found to rely on later processing stages in adults and the involvement of feedback connections (Bruns and Röder, 2010; Bonath et al., 2007). Emberson et al. (2015) provided evidence that the crossmodal connectivity is at least partially in place at the age of six months. Here we speculate that this initial crossmodal connectivity might even be more extensive in the developing brain (see Johannsen and Röder, 2014) and thus might be the neural underpinning of the enhanced sensitivity to simple crossmodal statistics in development which allows for quicker and a passive learning during infancy. We further assume in line with the ‘multisensory perceptual narrowing’ idea (Lewkowicz and Ghazanfar, 2006) that experience narrows down the initial crossmodal connectivity by eliminating non-confirmed connections while elaborating connections which are useful for an individual (Johannsen and Röder, 2014; Lewkowicz, 2014). These functionally tuned networks (including the experience dependent feedback connectivity) constitute models of the sensory world (Fiser et al., 2010). Their elaboration might go together with a switch towards model-based learning which is characterized by a larger context dependency. As some parts of the neural networks stabilize, learning must partially involve additional neural systems to guarantee that the adaptations necessary throughout life are realized without risking the loss of essential crossmodal knowledge. For example, prism wearing during the sensitive phase has been reported to change the connectivity between the central (ICC) and external (ICX) inferior colliculus of the auditory midbrain of barn owls. By contrast, crossmodal adaptation to prisms later in the critical period seems to be mediated by a reorganization of the optical tectum to which the ICX projects (Knudsen, 2002). Moreover, Bergan et al. (2005) reported that crossmodal adaptions to prims were enhanced in adult owls when they were allowed to hunt, that is, when such adaptations were particular needed. In accord with these findings in owls, we demonstrated that adult learning of crossmodal combinations depended on task relevance (Experiment 2b). Thus, as suggested by Keuroghlian and Knudsen (2007) and Bavelier et al. (2010), neuroplasticity in adults seem to require to a larger extend attention and behavioral relevance and thus the involvement of additional higher order neural systems. Task relevance or attention constitute specific top-down influences on sensory representations and are thus mediated via the feedback connections which become progressively tuned and elaborated during development (Dehaene-Lambertz and Spelke, 2015).

The present study was able to dissociate the processes for the learning of probabilities of sensory elements and for the learning of conditional probabilities of the sensory elements comprising crossmodal stimuli. All groups were sensitive to ‘Rare deviant stimuli’. To detect ‘Rare deviant stimuli’ the frequency of sensory elements rather than conditional probabilities had to be traced. Indeed, it was possible to detect ‘Rare deviant stimuli’ only based on one of the two sensory elements. The ERP effect to ‘Rare deviant stimuli’ started earlier than the ERP effects to ‘Rare recombined stimuli’. Such early deviant effects are typical for an auditory mismatch negativity (MMN, see Schröger and Wolff, 1996; Cheour et al., 2000). Therefore, we suggest that the observed ‘Rare deviant stimuli’ effect, similarly as has been proposed for the MMN reflects, indicates a sensory memory trace, which represents the frequency of sensory elements (Näätänen and Alho, 1995). By contrast, the detection of conditional probabilities of crossmodal stimuli cannot be based on such (unisensory) sensory memory traces. Thus, we speculate that the detection of ‘Rare deviant stimuli’, is based on modality specific systems (Frost et al., 2015). Although it has been reported that auditory mismatch responses are enhanced by redundant crossmodal (somatosensory) information in adults such a multisensory enhancement was only observed for later time epochs (>200 ms; Butler et al., 2012) than the first ‘Rare deviant stimulus’ effect of the present study.

Since we argue that the change in learning mode during development is related to functional specialization, the strong lateralization of both ERP effects in infants seems rather surprising. The differentiation of ‘Frequent standard stimuli’ and ‘Rare recombined stimuli’ requires the detection of conditional probabilities. This ability has been postulated as a precursor of language learning (Saffran et al., 1996). Indeed, it has been shown with structural imaging techniques that many hemispheric asymmetries, in particular those related to the language system (Friederici, 2009), exist at birth or shortly thereafter (see Dehaene-Lambertz and Spelke, 2015). Thus, we speculate that the strong left lateralized ERP difference between ‘Frequent standard stimuli’ and ‘Rare recombined stimuli’ might reflects a recruitment of similar neural circuits that have been proposed to enable the detection of word boundaries (Saffran et al., 1996), non-adjacent transitional probabilities and possibly syntactical rules (Friederici, 2002; Friederici et al., 2006). Thus, this neural system might, partially independently of sensory modality and domain, allow for detecting any type of statistical relations (Kuhl, 2010; Aslin and Newport, 2014). In fact, a correlation of syntactic competence and statistical learning skills in children has been reported (Kidd and Arciuli, 2016). The right lateralized ERP effect to ‘Rare deviant stimuli’ was not unique to the infant group, but was as well observed in the adults tested with the same passive design (Experiment 1b). Interestingly such a lateralization was neither found for Experiment 2a nor for Experiment 2b, in which the adult participants were actively engaged in a task. We speculate that ‘Rare deviant stimuli’ elicited a reflexive and exogenous attention shift to the rare sensory features. Such reflexive spatial attention orienting has often been associated with right parietal brain regions (Okada et al., 2008; Mort et al., 2003; Chica et al., 2011). In contrast, in Experiment 2a and 2b, participants had to allocate attention to a specific stimulus or stimulus combination and it was adaptive to avoid exogenous attention shifts.

In the present study ERP effects in adults were of different polarity and had a shorter latency compared to the infant group. We linked ERP effects in infants and adults based on the experimental manipulations and their relative timing. Due to the immature brain (e.g. incomplete myelination) of infants and children it is a common finding that absolute latencies of ERP effects are longer in the developing brain. Moreover, it has repeatedly been reported that polarities of effects differ in infancy or children and adulthood (Kouider et al., 2015; Neville et al., 2013; Nelson, 1997; de Haan and Halit, 2001).

In conclusion, our study demonstrates that six-month old infants were able to quickly learn crossmodal statistics through a mere passive exposure, whereas adults learned the same crossmodal combinations only when they were task relevant. Thus, we provide first evidence for a higher sensitivity for crossmodal statistics in infants compared to adults, indicating age-dependent mechanisms for the learning of arbitrary crossmodal combinations. We speculate that initial passive association learning allows infants to quickly form first internal models of their sensory environment. In adulthood these internal models are adjusted if this is behavioral adaptive.

1. 1. Arciuli J
   2. Simpson IC

   (2011) Statistical learning in typically developing children: the role of age and speed of stimulus presentation

   Developmental Science 14:464–473.

   https://doi.org/10.1111/j.1467-7687.2009.00937.x

   • PubMed
   • Google Scholar
2. 1. Aslin RN
   2. Newport EL

   (2014) Distributional language learning: Mechanisms and models of ategory formation

   Language Learning 64:86–105.

   https://doi.org/10.1111/lang.12074

   • PubMed
   • Google Scholar
3. 1. Aslin RN

   (2014) Infant learning: Historical, conceptual, and methodological challenges

   Infancy 19:2–27.

   https://doi.org/10.1111/infa.12036

   • PubMed
   • Google Scholar
4. 1. Bahrick LE
   2. Hernandez-Reif M
   3. Flom R

   (2005) The development of infant learning about specific face-voice relations

   Developmental Psychology 41:541–552.

   https://doi.org/10.1037/0012-1649.41.3.541

   • PubMed
   • Google Scholar
5. 1. Bahrick LE
   2. Lickliter R
   3. Flom R

   (2004) Intersensory redundancy guides the development of selective attention, perception, and cognition in infancy

   Current Directions in Psychological Science 13:99–102.

   https://doi.org/10.1111/j.0963-7214.2004.00283.x

   • Google Scholar
6. 1. Bahrick LE
   2. Lickliter R

   (2000) Intersensory redundancy guides attentional selectivity and perceptual learning in infancy

   Developmental Psychology 36:190–201.

   https://doi.org/10.1037/0012-1649.36.2.190

   • PubMed
   • Google Scholar
7. 1. Bavelier D
   2. Levi DM
   3. Li RW
   4. Dan Y
   5. Hensch TK

   (2010) Removing brakes on adult brain plasticity: from molecular to behavioral interventions

   Journal of Neuroscience 30:14964–14971.

   https://doi.org/10.1523/JNEUROSCI.4812-10.2010

   • PubMed
   • Google Scholar
8. 1. Bergan JF
   2. Ro P
   3. Ro D
   4. Knudsen EI

   (2005) Hunting increases adaptive auditory map plasticity in adult barn owls

   Journal of Neuroscience 25:9816–9820.

   https://doi.org/10.1523/JNEUROSCI.2533-05.2005

   • PubMed
   • Google Scholar
9. 1. Bonath B
   2. Noesselt T
   3. Martinez A
   4. Mishra J
   5. Schwiecker K
   6. Heinze HJ
   7. Hillyard SA

   (2007) Neural basis of the ventriloquist illusion

   Current Biology 17:1697–1703.

   https://doi.org/10.1016/j.cub.2007.08.050

   • PubMed
   • Google Scholar
10. 1. Brookes H
    2. Slater A
    3. Quinn PC
    4. Lewkowicz DJ
    5. Hayes R
    6. Brown E

    (2001) Three-month-old infants learn arbitrary auditory-visual pairings between voices and faces

    Infant and Child Development 10:75–82.

    https://doi.org/10.1002/icd.249

    • Google Scholar
11. 1. Bruns P
    2. Röder B

    (2010) Tactile capture of auditory localization: an event-related potential study

    European Journal of Neuroscience 31:1844–1857.

    https://doi.org/10.1111/j.1460-9568.2010.07232.x

    • PubMed
    • Google Scholar
12. 1. Bulf H
    2. Johnson SP
    3. Valenza E

    (2011) Visual statistical learning in the newborn infant

    Cognition 121:127–132.

    https://doi.org/10.1016/j.cognition.2011.06.010

    • PubMed
    • Google Scholar
13. 1. Butler JS
    2. Foxe JJ
    3. Fiebelkorn IC
    4. Mercier MR
    5. Molholm S

    (2012) Multisensory representation of frequency across audition and touch: high density electrical mapping reveals early sensory-perceptual coupling

    Journal of Neuroscience 32:15338–15344.

    https://doi.org/10.1523/JNEUROSCI.1796-12.2012

    • PubMed
    • Google Scholar
14. 1. Cheour M
    2. Leppänen PH
    3. Kraus N

    (2000) Mismatch negativity (MMN) as a tool for investigating auditory discrimination and sensory memory in infants and children

    Clinical Neurophysiology 111:4–16.

    https://doi.org/10.1016/S1388-2457(99)00191-1

    • PubMed
    • Google Scholar
15. 1. Chica AB
    2. Bartolomeo P
    3. Valero-Cabré A

    (2011) Dorsal and ventral parietal contributions to spatial orienting in the human brain

    Journal of Neuroscience 31:8143–8149.

    https://doi.org/10.1523/JNEUROSCI.5463-10.2010

    • PubMed
    • Google Scholar
16. Book

    1. de Haan M
    2. Halit H

    (2001)

    The development and neural basis of face processing during infancy

    In: Kalverboer A. F, Gramsbergen A, editors. Brain and Behavior in Human Development: A Sourcebook. Dordrecht: Kluwer Academic Publishers. pp. 921–937.

    • Google Scholar
17. 1. de Villers-Sidani E
    2. Chang EF
    3. Bao S
    4. Merzenich MM

    (2007) Critical period window for spectral tuning defined in the primary auditory cortex (A1) in the rat

    Journal of Neuroscience 27:180–189.

    https://doi.org/10.1523/JNEUROSCI.3227-06.2007

    • PubMed
    • Google Scholar
18. Book

    1. DeBoer T
    2. Scott LS
    3. Nelson CA

    (2007)

    Methods for acquiring and analyzing infant event-related potentials

    In: DeHaan M, editors. Infant EEG and Event-Related Potentials. New York: Psychology Press. pp. 5–38.

    • Google Scholar
19. 1. Dehaene-Lambertz G
    2. Spelke ES

    (2015) The infancy of the human brain

    Neuron 88:93–109.

    https://doi.org/10.1016/j.neuron.2015.09.026

    • PubMed
    • Google Scholar
20. 1. Emberson LL
    2. Conway CM
    3. Christiansen MH

    (2011) Timing is everything: changes in presentation rate have opposite effects on auditory and visual implicit statistical learning

    The Quarterly Journal of Experimental Psychology 64:1021–1040.

    https://doi.org/10.1080/17470218.2010.538972

    • PubMed
    • Google Scholar
21. 1. Emberson LL
    2. Richards JE
    3. Aslin RN

    (2015) Top-down modulation in the infant brain: Learning-induced expectations rapidly affect the sensory cortex at 6 months

    PNAS 112:9585–9590.

    https://doi.org/10.1073/pnas.1510343112

    • PubMed
    • Google Scholar
22. 1. Fantz RL

    (1964) Visual experience in infants: Decreased attention to familiar patterns relative to novel ones

    Science 146:668–670.

    https://doi.org/10.1126/science.146.3644.668

    • PubMed
    • Google Scholar
23. 1. Fiser J
    2. Aslin RN

    (2002) Statistical learning of new visual feature combinations by infants

    PNAS 99:15822–15826.

    https://doi.org/10.1073/pnas.232472899

    • PubMed
    • Google Scholar
24. 1. Fiser J
    2. Berkes P
    3. Orbán G
    4. Lengyel M

    (2010) Statistically optimal perception and learning: from behavior to neural representations

    Trends in Cognitive Sciences 14:119–130.

    https://doi.org/10.1016/j.tics.2010.01.003

    • Google Scholar
25. 1. Fletcher J
    2. Maybery MT
    3. Bennett S

    (2000) Implicit learning differences: a question of developmental level?

    Journal of Experimental Psychology: Learning, Memory, and Cognition 26:246–252.

    https://doi.org/10.1037/0278-7393.26.1.246

    • PubMed
    • Google Scholar
26. 1. Frank MC
    2. Slemmer JA
    3. Marcus GF
    4. Johnson SP

    (2009) Information from multiple modalities helps 5-month-olds learn abstract rules

    Developmental Science 12:504–509.

    https://doi.org/10.1111/j.1467-7687.2008.00794.x

    • PubMed
    • Google Scholar
27. 1. Friederici AD
    2. Bahlmann J
    3. Heim S
    4. Schubotz RI
    5. Anwander A

    (2006) The brain differentiates human and non-human grammars: functional localization and structural connectivity

    PNAS 103:2458–2463.

    https://doi.org/10.1073/pnas.0509389103

    • PubMed
    • Google Scholar
28. 1. Friederici AD

    (2002) Towards a neural basis of auditory sentence processing

    Trends in Cognitive Sciences 6:78–84.

    https://doi.org/10.1016/S1364-6613(00)01839-8

    • PubMed
    • Google Scholar
29. 1. Friederici AD

    (2009) Pathways to language: fiber tracts in the human brain

    Trends in Cognitive Sciences 13:175–181.

    https://doi.org/10.1016/j.tics.2009.01.001

    • PubMed
    • Google Scholar
30. 1. Frost R
    2. Armstrong BC
    3. Siegelman N
    4. Christiansen MH

    (2015) Domain generality versus modality specificity: the paradox of statistical learning

    Trends in Cognitive Sciences 19:117–125.

    https://doi.org/10.1016/j.tics.2014.12.010

    • Google Scholar
31. 1. Gómez R
    2. Maye J

    (2005) The developmental trajectory of nonadjacent dependency learning

    Infancy 7:183–206.

    https://doi.org/10.1207/s15327078in0702_4

    • Google Scholar
32. 1. Janacsek K
    2. Fiser J
    3. Nemeth D

    (2012) The best time to acquire new skills: age-related differences in implicit sequence learning across the human lifespan

    Developmental Science 15:496–505.

    https://doi.org/10.1111/j.1467-7687.2012.01150.x

    • PubMed
    • Google Scholar
33. 1. Johannsen J
    2. Röder B

    (2014) Uni- and crossmodal refractory period effects of event-related potentials provide insights into the development of multisensory processing

    Frontiers in Human Neuroscience 8:552.

    https://doi.org/10.3389/fnhum.2014.00552

    • PubMed
    • Google Scholar
34. 1. Jost E
    2. Conway CM
    3. Purdy JD
    4. Hendricks MA

    (2011)

    Neurophysiological correlates of visual statistical learning in adults and children

    Paper Presented at the 33rd Annual Meeting of the Cognitive Science Society 33:.

    • Google Scholar
35. 1. Keuroghlian AS
    2. Knudsen EI

    (2007) Adaptive auditory plasticity in developing and adult animals

    Progress in Neurobiology 82:109–121.

    https://doi.org/10.1016/j.pneurobio.2007.03.005

    • PubMed
    • Google Scholar
36. 1. Kidd E
    2. Arciuli J

    (2016) Individual differences in statistical learning predict children's comprehension of syntax

    Child Development 87:184–193.

    https://doi.org/10.1111/cdev.12461

    • PubMed
    • Google Scholar
37. 1. Kirkham NZ
    2. Slemmer JA
    3. Johnson SP

    (2002) Visual statistical learning in infancy: evidence for a domain general learning mechanism

    Cognition 83:B35–B42.

    https://doi.org/10.1016/S0010-0277(02)00004-5

    • Google Scholar
38. 1. Kirkham NZ
    2. Slemmer JA
    3. Richardson DC
    4. Johnson SP

    (2007) Location, location, location: development of spatiotemporal sequence learning in infancy

    Child Development 78:1559–1571.

    https://doi.org/10.1111/j.1467-8624.2007.01083.x

    • PubMed
    • Google Scholar
39. 1. Knudsen EI

    (2002) Instructed learning in the auditory localization pathway of the barn owl

    Nature 417:322–328.

    https://doi.org/10.1038/417322a

    • PubMed
    • Google Scholar
40. 1. Kouider S
    2. Long B
    3. Le Stanc L
    4. Charron S
    5. Fievet AC
    6. Barbosa LS
    7. Gelskov SV

    (2015) Neural dynamics of prediction and surprise in infants

    Nature Communications 6:8537.

    https://doi.org/10.1038/ncomms9537

    • PubMed
    • Google Scholar
41. 1. Krogh L
    2. Vlach HA
    3. Johnson SP

    (2012) Statistical learning across development: flexible yet constrained

    Frontiers in Psychology 3:598.

    https://doi.org/10.3389/fpsyg.2012.00598

    • PubMed
    • Google Scholar
42. 1. Kuhl PK

    (2010) Brain mechanisms in early language acquisition

    Neuron 67:713–727.

    https://doi.org/10.1016/j.neuron.2010.08.038

    • PubMed
    • Google Scholar
43. Book

    1. Lany J
    2. Saffran JR

    (2013)

    Statistical learning mechanisms in infancy

    In: Rubenstein J. L. R, Rakic P, editors. Comprehensive Developmental Neuroscience: Neural Circuit Development and Function in the Brain, Vol. 3. Amsterdam: Elsevier. pp. 231–248.

    • Google Scholar
44. 1. Lewkowicz DJ
    2. Ghazanfar AA

    (2006) The decline of cross-species intersensory perception in human infants

    PNAS 103:6771–6774.

    https://doi.org/10.1073/pnas.0602027103

    • PubMed
    • Google Scholar
45. Book

    1. Lewkowicz DJ
    2. Kraebel KS

    (2004)

    The value of multisensory redundancy in the development of intersensory perception

    In: Calvert G. A, Spence C, Stein B. E, editors. The Handbook of Multisensory Processes. Cambridge: MIT. pp. 655–678.

    • Google Scholar
46. 1. Lewkowicz DJ

    (1992) Infants' responsiveness to the auditory and visual attributes of a sounding/moving stimulus

    Perception & Psychophysics 52:519–528.

    https://doi.org/10.3758/BF03206713

    • PubMed
    • Google Scholar
47. 1. Lewkowicz DJ

    (2014) Early experience and multisensory perceptual narrowing

    Developmental Psychobiology 56:292–315.

    https://doi.org/10.1002/dev.21197

    • Google Scholar
48. 1. Maybery M
    2. Taylor M
    3. O'brien-Malone A

    (1995) Implicit learning: Sensitive to age but not IQ

    Australian Journal of Psychology 47:8–17.

    https://doi.org/10.1080/00049539508258763

    • Google Scholar
49. 1. Miller J

    (1982) Divided attention: evidence for coactivation with redundant signals

    Cognitive Psychology 14:247–279.

    https://doi.org/10.1016/0010-0285(82)90010-X

    • PubMed
    • Google Scholar
50. 1. Mort DJ
    2. Malhotra P
    3. Mannan SK
    4. Rorden C
    5. Pambakian A
    6. Kennard C
    7. Husain M

    (2003) The anatomy of visual neglect

    Brain 126:1986–1997.

    https://doi.org/10.1093/brain/awg200

    • PubMed
    • Google Scholar
51. 1. Näätänen R
    2. Alho K

    (1995) Mismatch negativity--a unique measure of sensory processing in audition

    International Journal of Neuroscience 80:317–337.

    https://doi.org/10.3109/00207459508986107

    • PubMed
    • Google Scholar
52. 1. Neil PA
    2. Chee-Ruiter C
    3. Scheier C
    4. Lewkowicz DJ
    5. Shimojo S

    (2006) Development of multisensory spatial integration and perception in humans

    Developmental Science 9:454–464.

    https://doi.org/10.1111/j.1467-7687.2006.00512.x

    • PubMed
    • Google Scholar
53. Book

    1. Nelson CA

    (1997)

    Electrophysiological correlates of memory development in the first year of life

    In: Reese H. W, Franzen M. D, editors. Biological and Neuropsychological Mechanisms. Life-Span Developmental Psychology. Mahwah: Lawrence Erlbaum Associates Inc. pp. 95–131.

    • Google Scholar
54. 1. Nemeth D
    2. Janacsek K
    3. Fiser J

    (2013) Age-dependent and coordinated shift in performance between implicit and explicit skill learning

    Frontiers in Computational Neuroscience 7:147.

    https://doi.org/10.3389/fncom.2013.00147

    • PubMed
    • Google Scholar
55. 1. Neville HJ
    2. Stevens C
    3. Pakulak E
    4. Bell TA
    5. Fanning J
    6. Klein S
    7. Isbell E

    (2013) Family-based training program improves brain function, cognition, and behavior in lower socioeconomic status preschoolers

    PNAS 110:12138–12143.

    https://doi.org/10.1073/pnas.1304437110

    • PubMed
    • Google Scholar
56. 1. Okada T
    2. Sato W
    3. Kubota Y
    4. Usui K
    5. Inoue Y
    6. Murai T
    7. Hayashi T
    8. Toichi M

    (2008) Involvement of medial temporal structures in reflexive attentional shift by gaze

    Social Cognitive and Affective Neuroscience 3:80–88.

    https://doi.org/10.1093/scan/nsm027

    • PubMed
    • Google Scholar
57. 1. Picton TW
    2. Bentin S
    3. Berg P
    4. Donchin E
    5. Hillyard SA
    6. Johnson R
    7. Miller GA
    8. Ritter W
    9. Ruchkin DS
    10. Rugg MD
    11. Taylor MJ

    (2000) Guidelines for using human event-related potentials to study cognition: recording standards and publication criteria

    Psychophysiology 37:127–152.

    https://doi.org/10.1111/1469-8986.3720127

    • PubMed
    • Google Scholar
58. 1. Riedel B
    2. Burton AM

    (2006) Auditory sequence learning: differential sensitivity to task relevant and task irrelevant sequences

    Psychological Research Psychologische Forschung 70:337–344.

    https://doi.org/10.1007/s00426-005-0226-9

    • PubMed
    • Google Scholar
59. 1. Saffran JR
    2. Aslin RN
    3. Newport EL

    (1996) Statistical learning by 8-month-old infants

    Science 274:1926–1928.

    https://doi.org/10.1126/science.274.5294.1926

    • PubMed
    • Google Scholar
60. 1. Saffran JR
    2. Johnson EK
    3. Aslin RN
    4. Newport EL

    (1999) Statistical learning of tone sequences by human infants and adults

    Cognition 70:27–52.

    https://doi.org/10.1016/S0010-0277(98)00075-4

    • PubMed
    • Google Scholar
61. 1. Scheier C
    2. Lewkowicz DJ
    3. Shimojo S

    (2003) Sound induces perceptual reorganization of an ambiguous motion display in human infants

    Developmental Science 6:233–241.

    https://doi.org/10.1111/1467-7687.00276

    • Google Scholar
62. 1. Schröger E
    2. Wolff C

    (1996) Mismatch response of the human brain to changes in sound location

    NeuroReport 7:3005–3008.

    https://doi.org/10.1097/00001756-199611250-00041

    • PubMed
    • Google Scholar
63. 1. Turk-Browne NB
    2. Jungé J
    3. Scholl BJ

    (2005) The automaticity of visual statistical learning

    Journal of Experimental Psychology: General 134:552–564.

    https://doi.org/10.1037/0096-3445.134.4.552

    • PubMed
    • Google Scholar
64. Book

    1. Zelazo PD
    2. Carlson SM
    3. Kesek A

    (2008)

    Development of executive function in childhood

    In: Nelson C. A, Luciana M, editors. Handbook of Developmental Cognitive Neuroscience. Cambridge: MIT Press. pp. 553–574.

    • Google Scholar

Author details

1. Sophie Rohlf

   Biological Psychology and Neuropsychology, University of Hamburg, Hamburg, Germany

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Visualization, Writing—original draft, Project administration

   Contributed equally with

   Boukje Habets

   For correspondence

   sophie.rohlf@uni-hamburg.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8947-5613

2. Boukje Habets

   1. Biological Psychology and Neuropsychology, University of Hamburg, Hamburg, Germany
   2. Biological Psychology and Cognitive Neuroscience, University of Bielefeld, Bielefeld, Germany

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Project administration, Writing—review and editing

   Contributed equally with

   Sophie Rohlf

   Competing interests

   No competing interests declared

3. Marco von Frieling

   Biological Psychology and Neuropsychology, University of Hamburg, Hamburg, Germany

   Contribution

   Conceptualization, Data curation, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

4. Brigitte Röder

   Biological Psychology and Neuropsychology, University of Hamburg, Hamburg, Germany

   Contribution

   Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Writing—original draft, Project administration

   Competing interests

   No competing interests declared

• 1,553

  views

• 226

  downloads

• 13

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.