Learning nonadjacent dependencies embedded in sentences of an artificial language: When learning breaks down.

Abstract

The structure of natural languages give rise to many dependencies in the linear sequences of words, and within words themselves. Detecting these dependencies is arguably critical for young children in learning the underlying structure of their language. There is considerable evidence that human adults and infants are sensitive to the statistical properties of sequentially adjacent items. However, the conditions under which learners detect nonadjacent dependencies (NADs) appears to be much more limited. This has resulted in proposals that the kinds of learning mechanisms learners deploy in processing adjacent dependencies are fundamentally different from those deployed in learning NADs. Here we challenge this view. In 4 experiments, we show that learning both kinds of dependencies is hindered in conditions when they are embedded in longer sequences of words, and facilitated when they are isolated by silences. We argue that the findings from the present study and prior research is consistent with a theory that similar mechanisms are deployed for adjacent and nonadjacent dependency learning, but that NAD learning is simply computationally more complex. Hence, in some situations NAD learning is only successful when constraining information is provided, but critically, that additional information benefits adjacent dependency learning in similar ways. (PsycINFO Database Record

Non‐adjacent Dependency Learning in Humans and Other Animals

“…Two recent studies using mixed complexity artificial grammars, which contain both adjacent and nonadjacent dependencies, showed that macaque monkeys were able to learn relationships between adjacent stimuli in the auditory or visual modality, but they found no evidence that they were sensitive to the nonadjacent dependencies (Milne et al., 2017; Wilson et al., 2015). Although humans have been reported to simultaneously learn both adjacent and nonadjacent dependencies within a mixed‐complexity grammar (Romberg & Saffran, 2013a; Wang & Mintz, 2018; Wilson et al., 2015) (although not in Milne et al., 2017), it is likely that in these studies the presence of salient adjacent relationships prevented the monkeys from learning these nonadjacent dependencies (as in infants in Gómez, 2002). However, a recent EEG experiment (Milne et al., 2016) reported that violations of nonadjacent dependencies evoked similar brain potentials in macaques as had previously been reported in humans using identical stimuli (Mueller et al., 2012).…”

“…This suggests that, while both adults and infants are able to implicitly learn nonadjacent dependencies, in some cases they may fail to do so if more salient adjacent cues are present. However, in some cases both adjacent and nonadjacent dependencies may be learned simultaneously (Romberg & Saffran, 2013b; Wang & Mintz, 2018), at least in some participants (see Milne, Petkov, & Wilson, 2017; Wilson et al., 2015), suggesting that the processes involved in learning these two types of relationships are not inherently antagonistic.…”

Learning non-adjacent rules and non-adjacent dependencies from human actions in 9-month-old infants

“…An additional property of many successful demonstrations of NAD learning in adults, such as [ 21 , 25 ], is that the trigrams with the NADs were presented as discrete, pre-segmented sequences, with 750ms of silence between each NAD trigram. When word sequences with the same statistical properties as in [ 21 ] were presented in a continuous sequences, adults did not learn the NADs [ 26 ]; similar learning failures in continuous sequences were found at the syllable level [ 28 ]. Some researchers even theorize that humans are constrained to learn NADs only when the non-adjacent items are at the edges of sequences, as defined by brief silences [ 29 ].…”

Individual differences in non-adjacent statistical dependency learning in infants

“…1 1 There is evidence from studies using ERP measures, rather than behavioral discrimination paradigms, that infants may even be able to track NADs with high variability in intervening elements by 3-4 months of age ( There is also evidence that the prosodic qualities of IDS may work in concert with statistical cues to facilitate learning NADs, just as they do in the case of phonetic tuning and word segmentation. For example, in artificial language materials NADs can be easier to learn when they are bracketed off from surrounding speech by pauses (Peña, Bonatti, Nespor, & Mehler, 2002;Wang & Mintz, 2018). In natural languages, NADs often bookend syntactic units, like noun phases and verb phrases, and IDS contains relatively long pauses at the edges of these syntactic units (Fisher & Tokura, 1996).…”

“…Here we focused on infants' ability to learn 'remote' or non-adjacent co-occurrence relationships (also called 'non-adjacent dependencies', or NADs), because they are pervasive in natural languages but can nonetheless be very hard to learn (e.g., Gómez, 2002, Newport & Aslin, 2004Wang & Mintz, 2018). An example of a such a relationship in English is the co-occurrence of the auxiliary is and the inflectional morpheme -ing that occurs on verbs, as in "She is playing".…”