Social Trait Information in Deep Convolutional Neural Networks Trained for Face Identification

Faces provide information about a person's identity, as well as their sex, age, and ethnicity. People also infer social and personality traits from the face — judgments that can have important societal and personal consequences. In recent years, deep convolutional neural networks (DCNNs) have proven adept at representing the identity of a face from images that vary widely in viewpoint, illumination, expression, and appearance. These algorithms are modeled on the primate visual cortex and consist of multiple processing layers of simulated neurons. Here, we examined whether a DCNN trained for face identification also retains a representation of the information in faces that supports social‐trait inferences. Participants rated male and female faces on a diverse set of 18 personality traits. Linear classifiers were trained with cross validation to predict human‐assigned trait ratings from the 512 dimensional representations of faces that emerged at the top‐layer of a DCNN trained for face identification. The network was trained with 494,414 images of 10,575 identities and consisted of seven layers and 19.8 million parameters. The top‐level DCNN features produced by the network predicted the human‐assigned social trait profiles with good accuracy. Human‐assigned ratings for the individual traits were also predicted accurately. We conclude that the face representations that emerge from DCNNs retain facial information that goes beyond the strict limits of their training.     

TRANSFER BETWEEN PICTURE NAMING AND TRANSLATION:

Abstract— A transfer paradigm was used to investigate the relationship between picture naming and translation English- Spanish bilinguals first named pictures and subsequently translated words in both their first (LI) and second (U) languages Some words in the translation task were repetitions of concepts that had previously been named as pictures Whereas picture naming produced reliable transfer to translation from LI to L2, It produced no transfer to translation from L2 to LI The results support the claim that connections in bilingual memory are asymmetric Translation is conceptually mediated from LI to L2 but lexically mediated from L2 to LI     

Functional ear (a)symmetry in brainstem neural activity relevant to encoding of voice pitch: a precursor for hemispheric specialization?

Pitch processing is lateralized to the right hemisphere; linguistic pitch is further mediated by left cortical areas. This experiment investigates whether ear asymmetries vary in brainstem representation of pitch depending on linguistic status. Brainstem frequency-following responses (FFRs) were elicited by monaural stimulation of the left and right ear of 15 native speakers of Mandarin Chinese using two synthetic speech stimuli that differ in linguistic status of tone. One represented a native lexical tone (Tone 2: T2); the other, T2′, a nonnative variant in which the pitch contour was a mirror image of T2 with the same starting and ending frequencies. Two 40-ms portions of f₀ contours were selected in order to compare two regions (R1, early; R2 late) differing in pitch acceleration rate and perceptual saliency. In R2, linguistic status effects revealed that T2 exhibited a larger degree of FFR rightward ear asymmetry as reflected in f₀ amplitude relative to T2′. Relative to midline (ear asymmetry = 0), the only ear asymmetry reaching significance was that favoring left ear stimulation elicited by T2′. By left- and right-ear stimulation separately, FFRs elicited by T2 were larger than T2′ in the right ear only. Within T2′, FFRs elicited by the earlier region were larger than the later in both ears. Within T2, no significant differences in FFRS were observed between regions in either ear. Collectively, these findings support the idea that origins of cortical processing preferences for perceptually-salient portions of pitch are rooted in early, preattentive stages of processing in the brainstem.     

Rural‐Urban Differences in Suicide Rates for Current Patients of a Public Mental Health Service in Australia

Rural versus urban rates of suicide in current patients of a large area mental health service in Australia were compared. Suicide deaths were identified from compulsory root cause analyses of deaths, 2003–2007. Age‐standardized rates of suicide were calculated for rural versus urban mental health service and compared using variance of age‐standardized rates with 95% confidence intervals. There were 44 suicides and the majority (62%) were rural. Only urban patients used jumping from heights as a method of suicide (4/17; p = 0.02). Rural patients had 2.7 times higher rates of suicide, similar to findings for rural versus urban community suicides and may reflect the underlying community rates, differences in mental health service delivery, or socioeconomic disadvantage.