Generalizing prosodic patterns by a non-vocal learning mammal

Juan M Toro,Marisa Hoeschele

doi:10.1007/s10071-016-1036-8

Abstract

Prosody, a salient aspect of speech that includes rhythm and intonation, has been shown to help infants acquire some aspects of syntax. Recent studies have shown that birds of two vocal learning species are able to categorize human speech stimuli based on prosody. In the current study, we found that the non-vocal learning rat could also discriminate human speech stimuli based on prosody. Not only that, but rats were able to generalize to novel stimuli they had not been trained with, which suggests that they had not simply memorized the properties of individual stimuli, but learned a prosodic rule. When tested with stimuli with either one or three out of the four prosodic cues removed, the rats did poorly, suggesting that all cues were necessary for the rats to solve the task. This result is in contrast to results with humans and budgerigars, both of which had previously been studied using the same paradigm. Humans and budgerigars both learned the task and generalized to novel items, but were also able to solve the task with some of the cues removed. In conclusion, rats appear to have some of the perceptual abilities necessary to generalize prosodic patterns, in a similar though not identical way to the vocal learning species that have been studied.

Highlights

From an early age, human infants take advantage of several sources of information to infer linguistic structure
If prosody processing is dependent on specialized traits, such as learning speciesspecific vocalizations, we should not observe it in mammals like rats, in which there is no evidence of vocal learning (e.g., Litvin et al 2007)
In contrast to what has been observed in budgerigars and humans in the same task (Hoeschele and Fitch 2016), rats failed to discriminate among test items when we removed individual acoustic cues including pitch, duration, intensity, and vowel quality

Summary

Introduction

Human infants take advantage of several sources of information to infer linguistic structure. Statistical regularities are only one of the cues infants might use to infer different aspects of language structure (e.g., Yang 2004) Another prominent source of information that is readily available in the signal is prosody. Tests with novel words demonstrated that they could generalize this discrimination to items they had not heard before This suggests the birds’ performance was based on categorical perception of stress patterns and not merely on the memorization of the specific training sequences. Two species of avian vocal learners (zebra finches and budgerigars) display a remarkable ability to detect and use prosody in speech as a differentiating cue much like humans do. Is the ability to use prosodic cues to discriminate between sequences widely shared across species distant in the phylogenetic tree, as it is for the ability to process statistical dependencies? Using the same methodology allowed us to compare a nonvocal learner’s performance directly with that of two vocal learning species

Methods

Procedure

Results

Discussion

Compliance with ethical standards