When control words lose control: emergence of top-down processing effects in Spanish spoken word recognition

Research on cross-dialectal spoken word recognition suggests integrated processing of social and linguistic information, enabled by the encoding of fine-grained sociophonetic detail. Asymmetries in encoding strength have also been documented, especially concerning idealized or socially meaningful variants. To date, the literature has demonstrated discrepancies between immediate and long-term processing, the latter being understudied, along with an overwhelming focus on English varieties. The current study therefore examined immediate and long-term processing of three /s/ variants in Peninsular Spanish: [s̺], [s̪], and [s̪^θ]. Each variant represents distinct geographic and social associations, providing a fruitful context to assess the implicit interaction of linguistic and extralinguistic information in processing. Seville capital participants (N = 42) completed a long-term form priming task containing word-initial /s/ items. Talkers (N = 6; one man and one woman per location) were from Seville capital, Seville outskirts, or central Spain. Response times initially demonstrated recognition equivalence in terms of processing speed: response times showed no differences among the three word-initial /s/ variants. However, faster processing of /s/ words produced by talkers from Seville capital and central Spain emerged after a long lag. Moreover, no recognition differences were observed between the two variants. Participants were less accurate in recognizing /s/ words produced with the Seville outskirts variant, suggesting processing difficulty. Unexpectedly, control words produced by one Seville capital talker were recognized especially quickly. The findings are discussed with respect to memory inequality and top-down processing effects and support the interaction of social and linguistic information in spoken word recognition.