Do perception verbs have a similar function as evidential inflections? A corpus-based study of English and Tibetan

International audience This paper examines whether perception verbs are used for the same function as grammaticalized evidentials. It is based on a contrastive corpus of Lhasa Tibetan (TSC; 4 pairs of Lhasa Tibetan native speakers, approx. 4 h. / 26,000 words) and English (CSC/LAC; 11 pairs of Standard British and American English native speakers, approx. 4 h. / 39,500 words). Tibetan possesses a multi-term evidential paradigm which prevails when referring to perceptions, while the English rendering of this domain is mainly lexical (Tournadre 2017, Whitt 2011, Mélac 2022). I will examine how inflections may compete with perception verbs in Lhasa Tibetan, and what may be the limits of the largely lexical system of English. First, the corpus indicates that Tibetan speakers refer to perception on average 7 times more frequently than English speakers (Mélac 2023). This is mainly due to the high frequency of direct perception inflections in Tibetan (42,846 instances per million words), while English speakers rarely specify whether they have directly perceived the event they are referring to (1,579 per millon words). Secondly, because the expression of perception is more lexical in English, speakers are compelled to be more specific when referring to this semantic domain. A grammatical morpheme such as Lhasa Tibetan ’dug may refer to all the senses as long as the speaker has directly perceived the event. The English equivalents of ’dug will usually involve verbs denoting a specific sensory channel, such as ‘see’ for visual sensations, ‘hear’ for auditory sensations, etc. Thirdly, the existence of grammatical and lexical evidentials in Tibetan provides more options in terms of informational status (Boye & Harder 2012, author XX, Boye 2023). Tibetan speakers may use a perception verb to draw attention to their sensations or an evidential inflection to make the perceptual aspect of the situation less noticeable. On average, the tools available in English to encode perception are more foregrounded, and do not offer such a range of attentional prominence for several reasons. However, we will see that Lhasa Tibetan does not seem to display the lexical richness of English when referring to perception. English has developed constructions that are absent in Tibetan, such as the various complementation of perception verbs (e.g., see + finite clause / + gerund / + bare infinitive clause), or the series of perceptual copulas (look / sound / feel / taste / smell + adjectival phrase). All these contrastive data confirm that, although languages develop profoundly different ways to encode perception, this semantic domain deeply impacts their structures, whether they possess a grammatical evidential paradigm or not.