TY - CONF

T1 - Strong competitors facilitate target name retrieval in simple picture naming

AU - Oppenheim, Gary

PY - 2017/9

Y1 - 2017/9

N2 - Intro: Semantic ‘inhibition’ in paradigms like picture-word interference is commonly assumed to reveal core properties of typical word production mechanisms: the distractor word cat interferes with naming a picture as dog because a lexical selection algorithm requires dog to overcome cat’s activation, so when cat is more activated it takes longer to select dog. However, considerable research over the past decade has raised questions about whether such ‘competitive’ RT effects may merely reflect artefacts of particular experimental tasks. Converging evidence from simpler tasks that lack such obvious experimental manipulations – such as timed picture-naming norms – would therefore strengthen the case that competition is an important, defining feature of typical lexical selection. In norms, name distributions for each picture are typically assumed to reflect the lexical activations on which selection operates, and pictures with higher name agreement are typically named faster than those with lower name agreement. Typical competitive selection accounts further predict that, ceteris paribus, concentrated competition from strong alternatives should hinder dominant name retrieval more than diffuse competition from an array of weaker competitors. For instance, given a picture that 50 people out of 100 name as truck, selecting truck should be slower if the remaining responses are split <45,5> between lorry and van (indicating a competitor nearly as strong as the target) than if they were split <25,25>. Methods: After collecting timed naming norms from 100 native UK English speaking students for the 525 black and white line drawings of the International Picture Naming Project (Bates et al, 2003), via standard norming procedures (ibid), for each picture I identified the dominant (i.e. most common) and secondary (i.e. second-most-common) names and their observed frequencies. Results: For 246 items, the dominant and secondary names were identified as coordinates of the same semantic category at the same level of specificity. These included synonyms and near-synonyms, e.g. truck/lorry, turtle/tortoise, but conservatively excluded simple morphophonological reductions or extensions, e.g. plane/aeroplane, hippopotamus/hippo, which might be considered subforms of a single lemma. Thus, for these items, the secondary name clearly represents a competitor for the dominant. Then, restricting the dataset to just the 18,516 trials in which participants produced the dominant name for these items, linear mixed effects regressions predicted their naming latencies as a function of the observed frequencies of both the picture’s a.) dominant name and b.) secondary name. Remarkably, increases in each predictor were associated with faster naming latencies. Observing faster dominant name RTs when the dominant name emerged more frequently replicates previous demonstrations that higher name agreement facilitates picture naming. But observing faster dominant name RTs when the secondary name emerged more frequently provides a novel challenge for theoretical claims that strong competitors should delay target word retrieval via competitive selection mechanisms.

AB - Intro: Semantic ‘inhibition’ in paradigms like picture-word interference is commonly assumed to reveal core properties of typical word production mechanisms: the distractor word cat interferes with naming a picture as dog because a lexical selection algorithm requires dog to overcome cat’s activation, so when cat is more activated it takes longer to select dog. However, considerable research over the past decade has raised questions about whether such ‘competitive’ RT effects may merely reflect artefacts of particular experimental tasks. Converging evidence from simpler tasks that lack such obvious experimental manipulations – such as timed picture-naming norms – would therefore strengthen the case that competition is an important, defining feature of typical lexical selection. In norms, name distributions for each picture are typically assumed to reflect the lexical activations on which selection operates, and pictures with higher name agreement are typically named faster than those with lower name agreement. Typical competitive selection accounts further predict that, ceteris paribus, concentrated competition from strong alternatives should hinder dominant name retrieval more than diffuse competition from an array of weaker competitors. For instance, given a picture that 50 people out of 100 name as truck, selecting truck should be slower if the remaining responses are split <45,5> between lorry and van (indicating a competitor nearly as strong as the target) than if they were split <25,25>. Methods: After collecting timed naming norms from 100 native UK English speaking students for the 525 black and white line drawings of the International Picture Naming Project (Bates et al, 2003), via standard norming procedures (ibid), for each picture I identified the dominant (i.e. most common) and secondary (i.e. second-most-common) names and their observed frequencies. Results: For 246 items, the dominant and secondary names were identified as coordinates of the same semantic category at the same level of specificity. These included synonyms and near-synonyms, e.g. truck/lorry, turtle/tortoise, but conservatively excluded simple morphophonological reductions or extensions, e.g. plane/aeroplane, hippopotamus/hippo, which might be considered subforms of a single lemma. Thus, for these items, the secondary name clearly represents a competitor for the dominant. Then, restricting the dataset to just the 18,516 trials in which participants produced the dominant name for these items, linear mixed effects regressions predicted their naming latencies as a function of the observed frequencies of both the picture’s a.) dominant name and b.) secondary name. Remarkably, increases in each predictor were associated with faster naming latencies. Observing faster dominant name RTs when the dominant name emerged more frequently replicates previous demonstrations that higher name agreement facilitates picture naming. But observing faster dominant name RTs when the secondary name emerged more frequently provides a novel challenge for theoretical claims that strong competitors should delay target word retrieval via competitive selection mechanisms.

M3 - Abstract

T2 - Architectures and Mechanisms in Language Processing 2017

Y2 - 7 September 2017 through 9 September 2017

ER -