Auditory memory: A comparison between humans and starlings

Abstract

In this study, we compare the processing of acoustic signals in European starlings (Sturnus vulgaris) and in human listeners by observing the decay of short-term auditory memory in delayed non-matching-to-sample experiments. A series of identical “sample” stimuli and a final “test” stimulus were separated by variable delays (1 to 180.1 s). Subjects had to classify sample and test stimuli as being either the same or different. Test stimuli were pure tones that differed in a single signal feature, i.e., frequency, and song motifs that differed in multiple signal characteristics. We have tested several predictions concerning the memory performance of starlings and humans and we obtained the following outcome: (1) In contrast to our expectation, signal complexity had no effect. The overall analysis of the starling data did not show differences in memory performance for signals differing in single or multiple signal features. (2) Starling and human data supported the hypothesis that auditory memory impairs with increasing delay. This was also seen when interfering noise was added to the delay periods in an additional series with human subjects. (3) The starling data showed that the repetition of sample stimuli improved memory performance, compared to only a single presentation. Human memory performance, however, was similar for a single and for the repeated presentation of signals. (4) Differences in salience between sample and test stimuli were positively related to memory performance only for tonal stimuli but not for song motifs. Results are discussed with respect to a model based on signal detection theory and to requirements for the analysis of natural communication signals.

Introduction

The structure of vocal signals in songbirds has striking parallels to that of human speech, and both song and speech are acquired by vocal learning (e.g., Doupe and Kuhl, 1999). Like speech, birdsong consists of a sequence of sounds. The smallest unit of song is termed an element that may be analogous to the phoneme, the smallest unit of speech (Doupe and Kuhl, 1999). Song elements form syllables (units of sound separated by silent intervals) which in turn form song types or “motifs” (e.g., Adret-Hausberger and Jenkins, 1988, Eens et al., 1989, Eens et al., 1991b, Gentner and Hulse, 2000). The timing and succession of syllables and motifs follow species-specific rules (song syntax; Doupe and Kuhl, 1999). The evaluation of complex sequences of communication signals demands storing signal elements or signal characteristics in the auditory memory. In this study, we investigate whether the starling (Sturnus vulgaris) is a suitable animal model for the perception and processing of acoustic signals in humans. We test whether humans and songbirds have comparable memory performance and memory persistence times when tested under similar experimental conditions.

A signal feature for which both species are most sensitive in detecting a change is the frequency. The just noticeable difference for frequency is better than for other signal characteristics like duration or rate of amplitude modulation (e.g., Fay, 1988). We employed pure tones as physically simple signals that provide frequency information and that only differed in a single signal characteristics (i.e., a change in frequency). To probe the effect of signal complexity on auditory memory, we decided to employ starling song motifs that exhibit differences in multiple signal characteristics (e.g., peak frequency, amplitude and frequency modulations). The song motifs presented to the subjects were chosen to have salient differences in peak frequencies. Multiple signal characteristics would provide more acoustical cues than a single signal feature. Thus we tested the hypothesis that memory performance for complex signals is superior to performance for simple signals in all experimental conditions (signal complexity hypothesis).

Similar to two previous studies (Zokoll et al., 2007, Zokoll et al., in press) memory performance was evaluated in a delayed non-matching-to-sample (DNMTS) paradigm. The subjects' task was to compare sample and test stimuli separated by delays of various duration and to decide whether sample and test were matching or non-matching. Our hypothesis was that auditory memory should fade significantly with increasing delay (delay hypothesis). Auditory memory persistence times were estimated on the basis of forgetting functions, i.e., functions that describe the decay in report accuracy as a function of increasing delays (e.g., Cowan, 1984, Cowan, 1997, White, 2001). Since, at least in starlings, song motifs are repeated during song bouts, we tested the hypothesis that the repetition of sample stimuli improves memory performance, compared to only a single presentation (sample repetition hypothesis). Finally we tested the hypothesis that differences in salience between sample and test stimuli are positively related to memory performance (salience hypothesis).