Optimality Theory as a General Cognitive Architecture

It was exactly 25 years ago that Paul Smolensky introduced Harmony Theory (Smolensky, 1986), a framework that would pursue an exciting, but certainly not straight path through linguistics (namely, Optimality Theory) and other cognitive domains. The goal of this workshop is not so much to look back to this path, but rather to discuss its potential continuation(s). Soon after its publication, Optimality Theory (OT) (Prince & Smolensky, 1993/2004) became one of the most successful frameworks for linguistic research. The number of publications submitted to the Rutgers Optimality Archive (at http://roa.rutgers.edu) exceeded one thousand in November 2008, with a large number of OT-related publications never added to ROA. The older sister of OT, Harmony Grammar (HG) (Smolensky, 1986) has also been the object of a recent raise in interest, especially since the publication of The Harmonic Mind (Smolensky & Legendre, 2006). Their key idea is that the linguistic mapping between form and meaning, or between underlying form and surface form, is realized by optimizing an abstract function: a real-valued one in HG, and a vector-valued one in OT. By developing the Integrated Connectionist/Symbolic (ICS) Cognitive Architecture, Smolensky and his colleagues worked out the connection between linguistics and a general theory of the mind/brain in a hardly precedented manner. Their approach is probably significantly closer to mainstream computational cognitive science than much of contemporary theoretical linguistics. General-purpose cognitive architectures (ACT-R) have also been combined with OT (Misker & Anderson, 2003; Rij, Rijn, & Hendriks, 2010). Learnability is addressed by, among others, Tesar, Boersma and Magri. The authors of the The Harmonic Mind alluded to the possibility that ICS—that is, Optimality Theory and Harmony Grammar—may prove a useful and adequate model of much of (higher) cognition, including domains beyond language. Simultaneously, and probably independently of their remark, simply as a consequence of OT’s success in linguistics, a number of scholars have advanced Optimality Theoretic models for non-linguistic phenomena. The authors of these isolated attempts usually even did not know of each other. Constraints applied to traffic rules (Boersma, 1998, 2003) and to a Talmudic dilemma (Dresher, 1996) aim only at illustrating the OT formalism. Parker and Parker (2004) present an analysis of ethical decision making in a religious context, which is clearly a first step toward an OT-style account of a non-linguistic cognitive function, despite potential criticism related to the cognitive grounding of their constraints. Although not elaborating on the connection with Optimality Theory, the “take the best” heuristics of the ABC Research Group can also be seen as an OT/HG-style approach to higher cognition (Gigerenzer, Todd, & the ABC Research Group, 1999). (Compare the lexicographic decision rule of Coenen and Marewski (2009) to strict domination in OT; see the discussions about the relation between OT and the ABC Research Group’s heuristics by Smolensky and Legendre (2006, vol. 1, p. 41-42) and Bı́ró (2006, p. 225f).) Even more explicit is the wish to view OT as a general cognitive framework for (higher) cognition in the work of Douglas Jones on kinship terminology, as well as of Tamás Biró on religious rituals. Jones (2003, 2010) developed an Optimality Theoretic model for one of the classic topics in cultural anthropology, the cross-cultural typology of kinship systems. Beside the connection to cognition in general, he also embedded his analysis in an evolutionary psychological account. His most recent article in BBS generated a vivid discussion on the applicability of OT beyond linguistics among the authors of the numerous peer-reviews. Biró (in press) has employed Optimality Theory to account for food taboos and the dynamics of different types of rituals. By the latter, he attempted computationally to underpin current theories in the cognitive science of religion.