Language change across generations for robots using cognitive maps

Languages change over time, as new words are invented, old words are lost through disuse, and the meanings of existing words are altered. The processes behind language change include the culture of language acquisition and the mechanisms used for language learning. We examine the effects of language acquisition and learning, in particular the length of the learning period over generations of robots. The robots form spatial concepts related to places in an environment: toponyms (place names) and simple prepositions (distances and directions). The use of spatial concepts allows us to investigate different classes of words within a single domain that provides a clear method for evaluating word use between agents. The individual words used by the agents can change rapidly through the generations depending on the learning period of the language learners. When the learning period is sufficiently long that more words are retained than invented, the lexicon becomes more stable and successful. This research demonstrates that the rate of language change depends on learning periods and concept formation, and that the language transmission bottleneck reduces the retention of words that are part of large lexicons more than words that are part of small lexicons