# Conspecific alarm calls, but not food-associated calls, elicit affect-based and object-based mental representations in a bonobo (Pan paniscus)

## TL;DR

Bonobos respond to alarm calls with mental representations of both fear and threats, but not to food-related calls without additional context.

## Contribution

Demonstrates that alarm calls in bonobos elicit both affect-based and object-based mental representations.

## Key findings

- Kanzi matched alarm calls to 'scare' and 'snake' lexigrams at above chance levels.
- Food-associated calls were not matched to 'surprise' or 'food' targets.
- Environmental cues are needed to interpret food-associated calls in varied contexts.

## Abstract

Non-human vocalizations carrying information regarding external events have been likened to referential words and are thus integral for exploring the origins of linguistic reference. Previous research suggests receivers decode this referential information and some studies have indicated that such calls can, like in humans, evoke mental representations of the referent in receivers. However, the nature of these representations remains ambiguous. Specifically, whether calls elicit affect-based representations (e.g. signaller fear after alarm calls) or object-based representations (e.g. threats encountered by signallers after alarm calls), or both, in listeners remains untested. To investigate this, we conducted a match-to-sample task with a language-competent bonobo (Kanzi) asking him to match playbacks of conspecific alarm and food-associated calls to lexigrams representing either affect-based (‘scare’, ‘surprise’) or object-based (‘snake’, ‘food’) content. Kanzi matched alarm calls to ‘scare’ and ‘snake’ lexigrams at above chance levels regardless of caller familiarity but did not match food-associated calls to either ‘surprise’ or ‘food’ targets. We propose environmental cues are required to interpret food-associated calls that occur across a variety of contexts. These findings suggest bonobo alarm calls evoke object- and affect-based representations for Kanzi, indicating the mechanisms underlying the perception of non-human vocalizations may be more similar to those in language than previously thought.