(Zipf, 1936) – Predictability in context (Piantadosi, Tily, & Gibson, 2011; Mahowald et al., 2013) Accounts appealing to meaning – Not considered because language thought to be arbitrary (Saussure, 1916)
individual speakers over time leads to the same regularity emerging in the structure of the lexicon through memory errors. Productive complexity bias Lexical complexity bias r= 0.69 1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Word Length (characters) Complexity Rating
method for simulating language change (e.g., Kirby, Cornish, & Smith, 2008) Gave the labels generated by participants to a new set of participants Iterated for total of 7 generations 50 participants/generation
minimum number of character edits necessary to transform one string to another can à cat  can à calculator  Normalized edit distance: distance between actual vs. guessed word, normalized by length of longest
change system – Chains with greater cross-generational change in lexical forms tend to show an increase in complexity bias over time. – When errors are made, consistent with complexity bias Pressure to simplify suppresses complexity bias – Because task not communicative, insufficient pressure against compression But, why doesn’t it strengthen?
problem of linkage In an iterated learning paradigm, find: – Language becomes more stable over time – Memory errors lead to complexity bias – Change in bias may be related to memory demands and listener pressures