Ventrolateral Prefrontal Neurons of the Monkey Encode Visual Instructions and Motor Behaviors in the Same Pragmatic Format

Abstract

The lateral prefrontal cortex (LPF), because of its extended anatomical connections with other cortical and subcortical areas, has access to a wide set of information regarding both the internal state of the subject and the external world, that determine its involvement in a broad spectrum of sensorimotor and cognitive processes. The resulting multidimensional representations enable this cortical sector to produce flexible strategies to navigate into the complex, ever-changing social environment, exploiting contextual and motivational information for selecting and implementing appropriate behaviors, as well as for inhibiting unnecessary or inappropriate ones (Miller, 2000; Rozzi and Fogassi, 2017; Tanji and Hoshi, 2008). Monkey electrophysiology and human fMRI studies suggested that information processing for action planning becomes more abstract when moving along a caudal-to-rostral gradient in the frontal cortex (Badre and D’Esposito, 2007, 2009; Koechlin et al., 2003; Koechlin and Summerfield, 2007). Despite theoretical differences among the authors, there is general agreement that the mid-portion of LPF is involved not only in behavior selection but also in planning actions. These functions require a strict relation between the middle sector of LPF and the parieto-premotor circuits subserving sensorimotor transformations. Accordingly, anatomical studies in the monkey have indicated that this sector, in particular the part corresponding to areas 12r and 46v, is anatomically connected with the parieto-premotor circuits for grasping control (Barbas and Pandya, 1989; Borra et al., 2011; Cavada and Goldman-Rakic, 1989; Cipolloni and Pandya, 1999; Gerbella et al., 2013; Saleem et al., 2014), suggesting that this prefrontal region could be an additional node of the lateral grasping network, involved in the context-based control of motor goals (see Rizzolatti et al., 2014). In line with these anatomical data, previous work from our lab (Simone et al. 2015) demonstrated that the ventral part of the lateral prefrontal cortex (VLPF) contains movement-related neurons, active during grasping execution both when the behavior is instructed by abstract rules and in naturalistic situation (Simone et al., 2015). Although these processes necessarily require the generation of goals based on the current context, it is largely unknown how VLPF neurons prospectively encode the instructing stimuli in relation to behavioral demands, and what is the specific format of the underlying neural representations. To tackle this issue, in Study 1, we analyzed the temporal dynamics of the responses VLPF neurons recorded during a Visuo-motor task instructed by visual cues, in which the monkey had to observe real objects and, subsequently, to perform (Action condition) or refrain (Inaction condition) object-oriented grasping actions. Our data show that VLPF recorded sector contains neurons responding in different task phases, and that the neuronal population discharge is stronger in the Inaction condition when the instructing cue is presented, and in the Action condition in the subsequent phases, from object presentation to action execution. Decoding analyses performed on neuronal populations showed that the activity recorded during the initial phases of the task shares the same type of format with that recorded during the final phases, suggesting the pragmatic nature of this format and that instructions and goals are encoded by prefrontal neurons as predictions of the action outcome. In Study 2, we aimed at assessing whether prefrontal neurons visual responses exclusively depend on the visual properties of the observed stimuli or are modulated by their pragmatic features and by the environmental contingencies. To this purpose, we recorded neuronal activity of prefrontal neurons in a Visual task requiring the monkeys simply to observe images of objects on a monitor. The recording sessions were carried out in the same days as for Study 1, allowing us to compare the visual responses in the two tasks in the same neurons. Our results indicate that part of VLPF neurons respond specifically to one stimulus or to a small set of stimuli, but there is no indication of a “passive” categorical coding. The comparison of neural responses recorded in the Visual and the Visuo-Motor tasks indicates that the visual responses to objects are often modulated by the task demands, with the strongest discharge when the object is target of an action.

Altogether, the data of the two studies indicate that VLPF neurons encode sensory stimuli (e.g. instructing cues and real objects) in relation to the current individual intention, and we propose that VLPF sensory-related responses are encoded at the neural level in terms of their behavioural outcome (pragmatic hypothesis).