TY - CONF

T1 - The effect of contextual bias on categorical perception

T2 - A Danish and Norwegian crosslinguistic study

AU - Ishkhanyan, Byurakn

AU - Fusaroli, Riccardo

AU - Højen, Anders

AU - Johansson, Christer

AU - Bleses, Dorthe

AU - Tylén, Kristian

AU - Christiansen, Morten H.

PY - 2020

Y1 - 2020

N2 - Danish has unusual phonetics and phonology. For instance, vocoids outnumber contoids and syllabic reductions result in both higher ambiguity and higher variability in the word forms than in other Scandinavian languages (Basbøll, 2005). These features may in part be the reason why Danish children are delayed in acquiring their native language (Bleses, Basbøll & Vach, 2011). Adult native Danish speakers, however, do not seem to have trouble understanding each other. Therefore, they may have developed certain cognitive strategies to cope with the unusual Danish phonetics and phonology. For example, they may be relying on context (top-down processing) to a larger extent than acoustic-phonetic input (bottom-up processing) when presented with unclear speech (e.g., Borsky, Tuller & Shapiro, 1998). In the current study, we explore whether Danish speakers indeed put more weight on top-down processes than, for instance, Norwegian speakers. Norwegian has the advantage in the current context of being culturally, historically and linguistically closely related to Danish, yet having a rather different phonology. To test our hypotheses, we adapted a spoken word recognition paradigm to Danish and Norwegian. The paradigm involves categorical perception and has previously been used for English (Connine, Blasko & Hall, 1991; Szostak & Pitt, 2013). These studies have shown that context plays a larger role when the target words begin with an ambiguous phone on either the [d]-[t] or the [s]-[ʃ] continuum. This contextual bias was stronger when the target word was separated from the disambiguating word by a larger number of syllables. In our study, we expected that Danish speakers would rely on context to a larger extent than Norwegian speakers. Therefore, the contextual bias effect would also be pronounced, when the target word is separated from the disambiguating word by a larger number of syllables. For our study, we constructed sentences that were either biased towards the target words sendt (1) or tændt (spelled tent in Norwegian) (2).(1) Hun har sendt et (langt og virkelig fint) brev.(2) Hun har tændt et (langt og virkelig fint) vokslys. Further, we manipulated the onset of the target word to a five-step-continuum with three intermediate steps and a clear [s] at one end and a clear [ts] ([th] in Norwegian) at the other end. We further created sentences with a smaller (NEAR condition) and a larger (FAR condition) distance between the target and the disambiguating word. The participants were instructed to decide whether they heard sendt or tændt and to click on the corresponding word on the screen. The results showed that there was a bias effect across distances and languages. When integrating response times and responses in a single dependent variable through a drift diffusion model (Ratcliff, 2000), the bias effect was stronger in the NEAR condition than in the FAR condition in Norwegian but such an interaction was not present in Danish. Thus, there was also a bias x distance x language 3-way interaction present, indicating that Danish speakers rely on context equally as much in both NEAR and FAR conditions, whereas Norwegian speakers behave more like English speakers, shown in previous literature. 

AB - Danish has unusual phonetics and phonology. For instance, vocoids outnumber contoids and syllabic reductions result in both higher ambiguity and higher variability in the word forms than in other Scandinavian languages (Basbøll, 2005). These features may in part be the reason why Danish children are delayed in acquiring their native language (Bleses, Basbøll & Vach, 2011). Adult native Danish speakers, however, do not seem to have trouble understanding each other. Therefore, they may have developed certain cognitive strategies to cope with the unusual Danish phonetics and phonology. For example, they may be relying on context (top-down processing) to a larger extent than acoustic-phonetic input (bottom-up processing) when presented with unclear speech (e.g., Borsky, Tuller & Shapiro, 1998). In the current study, we explore whether Danish speakers indeed put more weight on top-down processes than, for instance, Norwegian speakers. Norwegian has the advantage in the current context of being culturally, historically and linguistically closely related to Danish, yet having a rather different phonology. To test our hypotheses, we adapted a spoken word recognition paradigm to Danish and Norwegian. The paradigm involves categorical perception and has previously been used for English (Connine, Blasko & Hall, 1991; Szostak & Pitt, 2013). These studies have shown that context plays a larger role when the target words begin with an ambiguous phone on either the [d]-[t] or the [s]-[ʃ] continuum. This contextual bias was stronger when the target word was separated from the disambiguating word by a larger number of syllables. In our study, we expected that Danish speakers would rely on context to a larger extent than Norwegian speakers. Therefore, the contextual bias effect would also be pronounced, when the target word is separated from the disambiguating word by a larger number of syllables. For our study, we constructed sentences that were either biased towards the target words sendt (1) or tændt (spelled tent in Norwegian) (2).(1) Hun har sendt et (langt og virkelig fint) brev.(2) Hun har tændt et (langt og virkelig fint) vokslys. Further, we manipulated the onset of the target word to a five-step-continuum with three intermediate steps and a clear [s] at one end and a clear [ts] ([th] in Norwegian) at the other end. We further created sentences with a smaller (NEAR condition) and a larger (FAR condition) distance between the target and the disambiguating word. The participants were instructed to decide whether they heard sendt or tændt and to click on the corresponding word on the screen. The results showed that there was a bias effect across distances and languages. When integrating response times and responses in a single dependent variable through a drift diffusion model (Ratcliff, 2000), the bias effect was stronger in the NEAR condition than in the FAR condition in Norwegian but such an interaction was not present in Danish. Thus, there was also a bias x distance x language 3-way interaction present, indicating that Danish speakers rely on context equally as much in both NEAR and FAR conditions, whereas Norwegian speakers behave more like English speakers, shown in previous literature. 

M3 - Paper

ER -