Editorial of E-Book on Action and Language Integration

such as the characteristics of the agents implied by sentences. The authors also explicitly discuss the implications of these studies for embodied robotics. Symes and colleagues present a cognitive psychology study on the integrating action and language through biased competition. This is based on previous psychological investigations that have demonstrated that planning an action biases visual processing, as in Symes et al.’s (2008) findings reporting faster target detection for a changing object amongst several non-changing objects. This new experimental study investigates how this effect might compare to, and indeed integrate with, effects of language cues. Using the same change-detection scenes as in Symes et al. (2008), two effective sources of bias are identified: (i) action primes, and (ii) language cues. For example, a sentence as “Start looking for a change in the larger objects” cues object size, and these successfully enhanced detection of size-congruent targets. Additional experiments explore the biases’ co-occurrence within the same task, such as action prime (participants plan a power or precision grasp) and a language (a sentence) cue preceding stimulus presentation. Experimental results support the authors’ predictions from the biased competition model by Desimone and Duncan (1995), in particular reliably stronger effects of language, and concurrent biasing effects that were mutually suppressive and additive. Greco and Caneva (2010) focus on compositional symbol grounding for motor patterns. They propose a new comparative experimental/simulative paradigm to study the learning of compositional grounded representations for motor patterns. In a psychology experiment, participants learn to associate non-sense arm motor patterns, performed in three different hand postures, with non-sense words. Two experimental conditions are carried out: (i) in the compositional condition, each pattern was associated with a two-word (verb–adverb) sentence; (ii) in the holistic condition, each pattern was associated with a unique word. Experimental results show that the compositional group achieved better results in naming motor patterns, especially for patterns where hand postures discrimination was relevant. In order to ascertain the differential effects of memory load and of systematic grounding, neural network simulations were also carried out. After a basic simulation reproducing the default participants’ performance, in some simulations the number of stimuli (motor patterns and words) was increased and the systematic association between words and patterns was disrupted, while keeping the same number of words and compositionality. Simulation results show that in both conditions the advantage for the compositional condition significantly increased. This indicates that the advantage for the compositional condition may be related to systematicity rather than to mere informational gain. Overall, both experimental and simulation data support the hypothesis of a shared action/language compositional motor representation. Increasing theoretical and experimental research on action and language processing in humans and animals clearly demonstrates the strict interaction and co-dependence between language and action. This has been extensively demonstrated in neuroscientific investigations (e.g., Rizzolatti and Arbib, 1998; Cappa and Perani, 2003; Pulvermuller, 2003), psychology experiments (e.g., Glenberg and Kaschak, 2002; Pecher and Zwaan, 2005; Barsalou, 2008), evolutionary psychology (e.g., Corballis, 2002), and computational modeling (e.g., Cangelosi and Parisi, 2004; Massera et al., 2007; Cangelosi, 2010). All these studies have important implication both for the understanding of the action basis of cognition in natural and artificial cognitive systems, as well as for the design of cognitive and communicative capabilities in robots (Cangelosi et al., 2010). The journal “Frontiers in Neurorobotics” published a collection of articles on the topic of action and language integration both in natural cognitive systems (e.g., humans and animals) and in artificial cognitive agents (robots and simulated agents). These articles are now collected in an e-book, for wider dissemination. This set of chapters provides an up to date overview of current advances in the grounding of language into sensorimotor knowledge. The first chapters primarily focus on experimental evidence from cognitive psychology (Symes et al., 2010), cognitive neuroscience studies (Borghi et al., 2010), and comparative experimental/simulation studies (Greco and Caneva, 2010). Two chapters then use neural network simulation for motor chains for sentence processing (Chersi et al., 2010) and a computational model of gaze planning in word recognition and reading (Ferro et al., 2010). Finally, four chapters use cognitive systems and robotics methodologies to investigate general principles of action–language grounding (Parisi, 2010), teleological representations of action and language for human–robot interaction experiments (Lallee et al., 2010), verbal and non-verbal communication in neurorobotics models (Bicho et al., 2010), and action bases of action words (Marocco et al., 2010).