Although syntactic priming is well studied and commonly assumed to involve implicit learning, the mechanisms behind this phenomenon are still under debate. We tested whether implicit learning of adjacent and non-adjacent sequences occurs in a non-linguistic, finger sequence task (Serial Reaction Time task), and if so, whether these implicitly-learned dependencies can cause syntactic priming in the linguistic domain. We followed the logic that exposure to statistical patterns in the SRT task may influence language users’ relative clause (RC) attachment biases, and trained participants on SRT sequences with adjacent or non-adjacent dependencies. Participants then wrote completions to relative clause fragments in a situation where they could opt for adjacent or non-adjacent linguistic structures. Participants successfully learned the adjacent and non-adjacent dependency implicitly during the SRT task, but, strikingly, their RC continuations did not exhibit priming effects. Implications for theories of syntactic priming and its relations to implicit learning are discussed.

Many studies of human sequential pattern learning demonstrate that learners detect adjacent and non-adjacent dependencies in many kinds of sequences. However, it is often assumed that the computational mechanisms behind extracting these dependencies are the same. We replicate the seminal finding that adults are capable of learning dependencies between non-adjacent words (Gómez, 2002). When we eliminate the positional information about the statistical structures by embedding the structure in phrases, learners can no longer learn the dependencies. Our methods allow us to study the learning mechanisms that are more representative of the patterns in natural languages, and show that when directly compared, adjacent and non-adjacent dependencies are not equally learnable. We suggest that learning non-adjacent dependencies in language involves a different computational mechanism from learning adjacent dependencies

Many grammatical dependencies in natural language involve elements that are not adjacent, such as between the subject and verb in the child always runs. To date, most experiments showing evidence of learning non- adjacent dependencies have used artificial languages in which the to-be-learned dependencies are presented in isolation by presenting the minimal sequences that contain the dependent elements. However, dependencies in natural language are not typically isolated in this way. In this study we exposed learners to non-adjacent dependencies in long sequences of words. We accelerated the speed of presentation and learners showed evidence for learning of non-adjacent dependencies. The previous pause-based positional mechanisms for learning of non-adjacent dependency are challenged.

Grammatical dependencies often involve elements that are not adjacent. However, most experiments in which non-adjacent dependencies are learned bracketed the dependent material with pauses, which is not how dependencies appear in natural language. Here we report successful learning of embedded NAD without pause bracketing. Instead, we induce learners to compute structure in an artificial language by entraining them through processing English sentences. We also found that learning becomes difficult when grammatical entrainment causes learners to compute boundaries that are misaligned with NAD structures. In sum, we demonstrated that grammatical entrainment can induce boundaries that can carry over to reveal structures in novel language materials, and this effect can be used to induce learning of non-adjacent dependencies.

The phenomenon of syntactic priming is well studied in the literature, but the mechanisms behind it are still under debate. In this study, we trained English-speaking participants in artificial language sequences with dependencies that are either adjacent or non-adjacent. The participants then wrote completions to relative clause (RC) fragments. We found that participants who learn non-adjacent dependencies in the artificial language, exhibit a bias to write high-attachment (non-adjacent) continuations for RCs, when compared to participants in a control condition who exhibit low-attachment (adjacent) biases in RCs. The implications for theories of syntactic priming and its relations to implicit learning are discussed.

Existing models of statistical learning involve computation ofconditional probabilities over discrete, categorical items in asequence. We propose an alternative view that learning occursthrough a process of tracking changes along physicaldimensions from one stimulus to the next within a “perceptualsimilarity space.” To test this alternative, we examined asituation where it is difficult or impossible to label stimuli inreal time, and where the two assumptions lead to conflictinghypotheses. We conducted two experiments in which humanparticipants passively listened to a familiarization sequence offrequency-modulated tones and were then asked to makefamiliarity judgments on a series of test bigrams. Behavioralresults were broadly consistent with a conceptualization oflearning as tracking trajectories through perceptual similarityspace. We also trained a neural network that codes stimuli asvalues along two continuous dimensions to predict the nextstimulus given the current stimulus, and show that it capturedkey features of the human data.