Developmental phonagnosia: Neural correlates and a behavioral marker Xiaokun Xu, Irving Biederman ⇑ , Bryan E. Shilowich, Sarah B. Herald, Ori Amir, Naomi E. Allen University of Southern California, United States article info Article history: Received 24 October 2014 Revised 12 June 2015 Accepted 16 June 2015 Keywords: Phonagnosia Voice recognition fMRI Ventromedial prefrontal cortex Voice discrimination Voice imagination Temporal voice area Voice individuating cues (VICs) White matter tracts Celebrity voice recognition Prosopagnosia Person identity node abstract A 20-year old female, AN, with no history of neurological events or detectable lesions, was markedly poorer than controls at identifying her most familiar celebrity voices. She was normal at face recognition and in discriminating which of two speakers uttered a particular sentence. She evidences normal fMRI sensitivity for human speech and non-speech sounds. AN, and two other phonagnosics, were unable to imagine the voices of highly familiar individuals. A region in the ventromedial prefrontal cortex (vmPFC) was differentially activated in controls when imagining familiar celebrity voices compared to imagining non-voice sounds. AN evidenced no differential activation in this area, which has been termed a person identity semantic system. Rather than a deficit in the representation of voice-individuating cues, AN may be unable to associate those cues to the identity of a familiar person. In this respect, the deficit in developmental phonagnosia may bear a striking parallel to developmental prosopagnosia. Ó 2015 Elsevier Inc. All rights reserved. 1. Introduction Social animals, almost by definition, need to individuate other members of their own species. Face recognition has garnered the lion’s share of how humans individuate others but in our evolutionary history, before lighting and caller ID, being able to dis- tinguish people at night or in a dense forest or jungle on the basis of their voices could be a critical adaptation. So perhaps it is not sur- prising that almost all of us are able to readily identify a familiar person solely from their voice, just as we are able to identify that person from a glance at their face. A deficit in voice identification has been termed ‘‘phonagnosia’’ (Van Lancker & Canter, 1982). Kreiman and Sidtis (2011) distinguish language from prosody. It is prosody that largely underlies our ability to individuate a speaker on the basis of his or her voice. The units of language include entities such as phonemes, morphemes, words, sentences, phrases, and discourse. There is considerable variation in the dimensions or units that have been defined in the study of prosody. Kreiman and Sidtis list 108 attributes such as accent, cadence, final lowering, intonation, key, phrasing, pitch, and stress pattern that have been considered as characteristics of prosody. Prosody has generally been regarded as emerging from supersegmental varia- tion, i.e., variation that would be greater than individual phonemes or syllables. But information facilitating speaker identification is also conveyed by sublexical units, such as phonemes or syllables, as when vowels, for example, signal vocal tract characteristics. Kreiman and Sidtis do allow sublexical variation under a more general definition of prosody but it does seem useful to preserve the standard characterization of prosody as referring to supersegmental variation. Given that there does not seem to be a generally accepted term to characterize all the cues that might be employed in individuating voices, we here adopt the term Voice-Individuating Cues (VICs) to refer to those aspects of speech that are not language (in Kreiman and Sidtis’ sense) but convey information relevant to the individuation of a voice even if they are aspects of sublexical units. The dissociation of VICs and lan- guage processing is supported by the remarkable finding of Van Lancker and Canter (1982) that severely aphasic individuals with no understanding of language or a capacity for language produc- tion could, nonetheless, readily identify who was speaking. These individuals suffered lesions to the left hemisphere. As in prosopagnosia, phonagnosia can be either ‘‘developmen- tal’’ (likely congenital), or ‘‘acquired,’’ typically through damage to the right parietal region, as a result of injury or stroke (Van Lancker, Kreiman, & Cummings, 1989). At the time of our testing, http://dx.doi.org/10.1016/j.bandl.2015.06.007 0093-934X/Ó 2015 Elsevier Inc. All rights reserved. ⇑ Corresponding author. E-mail address: bieder@usc.edu (I. Biederman). Brain & Language 149 (2015) 106–117 Contents lists available at ScienceDirect Brain & Language journal homepage: www.elsevier.com/locate/b&l