Phenotyping psychosis: room for neurocomputational and content-dependent cognitive endophenotypes?

INTRODUCTION The endophenotype research strategy aims at reducing complex clinical phenomena to reveal a more tractable mapping to underlying genes. Cognitive dysfunctions have been widely pursued as target endophenotype in schizophrenia. We critically discuss the promise and limitations of this approach. METHODS Relevant theoretical and empirical issues on genes and behaviour, neurocognitive structure and psychopathology were selectively reviewed and discussed. RESULTS Some important inherent limitations of the current cognitive endophenotype approach were identified. These include reliance on (1) classic neuropsychology; (2) deficit measures; and (3) a general information processing approach with the use of content-independent, neutral stimuli. As a result, many current cognitive endophenotypes are likely to overlap and converge with general cognitive impairments, which may be shared with other disorders. CONCLUSIONS We propose three novel directions for further psychosis endophenotype research: (1) in addition to such content-independent computational processes, which operate in a similar way regardless of the stimuli, it is important to consider the potential roles of "content-dependent endophenotypes", which operate on different stimuli in consistently different manners. Advances in cognitive studies suggest there may be evolutionarily important aspects of cognition which are content-dependent. We propose that both content-independent and content-dependent processes should be addressed in psychosis research. (2) In line with the emphasis on content, close attention should be paid to the study of "psychopathological endophenotypes" in addition to cognitive endophenotypes. (3) "Neurocomputational endophenotypes" may be defined by parsing cognitive processes into "subsystems" with specific computational processing algorithms and considering key computational parameters suggested from these models. These potential "neurocomputational endophenotypes" (such as neuronal noise, synaptic learning algorithms) are potentially intermediate variables located between the levels of cognition and neurobiology.