Mapping the semantic homunculus: A functional and behavioural analysis of overt semantic generation

Previous neuroanatomical research has shown that semantic processing of action-related language activates the premotor, motor, and sensory cortices somatotopically (e.g., Tettamanti et al., J Cognitive Neurosci. 2005;17(2): 273-281, using a listening task, and Hauk et al., Neuron. 2004;41:301-307 and Pulvermuller et al., Eur J Neurosci 2005;21:793-797; J Cognitive Neurosci 2005;17(6):884-892 using a silent reading task). We examined this somatotopic semantics hypothesis using an overt semantic generation task (i.e., participants generated aloud their own personal description of how they would interact with target object words), rather than semantic comprehension as examined in previous research, so as to provide a stronger test of the hypothesis under conditions that tap one's own semantic knowledge about interacting with objects. Experiment 1 used functional Magnetic Resonance Imaging (fMRI) to examine somatotopically organized activation in the premotor cortex for an overt semantic generation task, using targets that naturally involve either arm interactions or leg interactions. Consistent with previous research, our results showed that semantic processing related to object interaction involves the motor, premotor and sensory cortices in a somatotopic fashion. Previous behavioural research has shown a response advantage in lexical decision for words with multiple meanings or features, which diminishes with tasks that decrease semantic involvement (e.g., Borowsky and Masson, J Exp Psychol: Learn Memory Cognit 1996; 22(1):63-85; Pexman et al., Psychon Bull Rev 2002;9(3): 542-549). Experiment 2 evaluated whether semantic generation response times (total duration of response) display a complexity advantage (i.e., faster response times for more complex objects), and whether complexity ratings were related to the volume of brain activation during the task. Results from this behavioural experiment revealed a significant negative relationship between the total duration of response (i.e., the total amount of time taken to respond to the stimuli) and object complexity for leg objects (a semantic complexity advantage), but not for arms. This suggests that the smaller repertoire of possible interactions with leg objects requires a greater reliance on semantic knowledge in order to respond in the semantic generation task. This interpretation was further supported by a greater volume of brain activation in the premotor cortex for arm objects versus leg objects. The response times from Experiment 1 were also compared to the semantic complexity ratings gathered in Experiment 2 to determine if response times in the fMRI environment were affected by how complex an object is.