Miller's Monkey Updated: Communicative Efficiency and the Statistics of Words in Natural Language

Abstract

(In Press at Cognition). Is language designed for communicative and functional efficiency? G. K. Zipf provided a famous argument in the affirmative: shorter words, which are easier to use, tend to be more frequent, apparently due to the pressure to minimize speaker effort. However, G. A. Miller showed that even a monkey randomly typing at a keyboard, and intermittently striking the space bar, would generate "words" with similar statistical properties. Recent quantitative analyses of human language lexicons (Piantadosi, Tily, & Gibson, 2012) have revived Zipf's functionalist hypothesis. Ambiguous words tend to be short, frequent, and easier to articulate in language production, which is interpreted as evidence for communicative efficiency as the context of language use often provides cues to overcome lexical ambiguity. In this study, we update Miller's monkey thought experiment to incorporate empirically motivated phonological and semantic constraints on the creation of words. We claim that the appearance of communicative efficiency is a spandrel (Gould & Lewontin, 1979), as lexicons formed without the context of language use or reference to communication or efficiency exhibit comparable statistical properties. Furthermore, the updated monkey model provides a good fit for the growth trajectory of English as recorded in the Oxford English Dictionary. Focusing on the history of English words since 1900, we show that lexicons resulting from the monkey model provide a better embodiment of communicative efficiency than the actual lexicon of English. We conclude by arguing for the need to go beyond correlational statistics and to seek direct evidence for the mechanisms that underlie principles of language design.

The forms and meanings of grammatical markers support efficient communication

“…Our work suggests that systems of grammatical markers achieve efficient trade-offs between informativeness and simplicity, but does not capture the historical processes that led to this outcome. It is possible that efficient trade-offs could arise in the absence of communicative pressures (Caplan, Kodner, & Yang, 2020), but recent work on cultural evolution and language acquisition suggests that language learning and use impose pressures towards informativeness and simplicity (Kirby, Tamariz, Cornish, & Smith, 2015;Carstensen, Xu, Smith, & Regier, 2015;Carr, Smith, Culbertson, & Kirby, 2020). On this account, the pressure towards informativeness applies during cooperative language use (e.g., Fay, Garrod, Roberts, & Swoboda, 2010), and the pressure towards simplicity applies during language learning (e.g., Hudson Kam & Newport, 2005).…”

On Homophony and Rényi Entropy

“…Their results, though, are vulnerable to similar criticisms to the ones that we provide in this paper about the use of potentially overfit n-gram models. Caplan et al (2020) propose a phonotactic monkey-an extension of Miller's (1957) random typing thought experiment-to make a similar point to Trott and Bergen. They rely on similar n-gram phonotactic models for their analysis of homophony, being thus vulnerable to the same criticisms we present here.…”

Why do some words have more meanings than others? A true neutral model for the meaning-frequency correlation.

“…The interpretation of this result has recently been challenged by computational modelling (Caplan et al, 2020;Trott & Bergen, 2020). In these studies, the authors attempted to construct null models devoid of selection for efficient communication.…”

“…In the neutral model presented in this paper, words are chosen to be reused at random, independently of their frequency, hence there is no selection mechanism favoring efficient communication. Unlike previous attempts to introduce null models of the meaning-frequency correlation (Caplan, Kodner, & Yang, 2020;Trott & Bergen, 2020), it truly does not rely on selection for frequency. We show that statistical regularities related to ambiguity, such as Zipf's meaning-frequency correlation, can arise in conditions when words are not undergoing any selective pressures.…”