Efficient coding in dolphin surface behavioral patterns

W ords that are more frequently used tend to be shorter in human language [1, 2] [Figure 1(a)]. The length of a word can be measured in letters or phonemes. The shortening of words with frequency can be regarded as evidence of efficient coding: by employing shorter codes for more frequent words, it is possible to increase the rate of information transmitted [3]. This tendency is the rationale behind file compression techniques such as Huffman codes [4]. Efficient coding has not been reported in other species to our knowledge. Here, we will provide the first evidence of a negative correlation between frequency and code size in the surface behavioral patterns of bottlenose dolphins (Tursiops sp.). Surface behavioral patterns represent a series of body movement and behavioral units, which can be clearly distinguished as a bout at the water surface. Our standard classification of behavioral patterns has been used to define the ethogram of dolphin populations in many previous studies (see Ref. 5 for a review). Patterns, described in Ref. 5, were defined to be mutually exclusive and cumulatively inclusive. As a whole, they describe the entire behavioral repertoire of the bottlenose dolphin population living in Doubtful Sound, New Zealand, that can be observed at the surface. Here, we defined behavioral units that were unambiguously representing distinct parts in the body movement, which represented behavioral patterns. They were also defined to be mutually exclusive. We defined these units and pattern composition independently from this study and prior to it. Thus, this analysis did not influence the definition of units and patterns. These behavioral units can be used in different combinations to produce different behavioral patterns, which in many cases involve more than one behavioral unit (Table *Present address: Institute of Biological and Environmental Sciences, University of Aberdeen, Tillydrone Avenue, Aberdeen, AB24 2TZ, UK Correspondence to: Ramon Ferrer-i-Cancho, Departament de Llenguatges i Sistemes Informàtics, TALP Research Center, Universitat Politècnica de Catalunya, Campus Nord, Edifici B6 C/Jordi Girona Salgado 1-3, 08034 Barcelona, Spain. (e-mail: rferrericancho@lsi.upc.edu)