Posters: Language I

Semantic context effects in naming tasks have been most influential in devising and evaluating models of word production. In the present study we investigated semantic context effects in one prominent task, blocked-cyclic naming, in which stimuli are named repeatedly either sorted by semantic category (homogeneous context) or intermixed (heterogeneous context). Previous blocked-cyclic naming studies have consistently shown that picture naming responses are slowed down in the homogeneous context from the second cycle (i.e., presentation) onwards. We compared semantic context effects in two task versions, picture naming and sound naming. Target words were identical across task versions (e.g., participants responded with “dog” to either the picture of a dog or to the sound produced by a dog [barking]). We found that (a) semantic interference in the homogeneous context was also obtained with sounds and (b) that the effect was substantially larger with sounds than with pictures. This was true regardless of whether stimulus type was constant (tested between participants, Experiments 1 and 2) or mixed (tested within participants, Experiment 3). Our findings demonstrate that (a) semantic context effects in blocked-cyclic naming generalize to stimulus types other than pictures and (b) are modulated by pre-lexical processes that depend on the nature of the stimuli used for eliciting the naming responses.

Studies have demonstrated that listeners can retain detailed voice-specific acoustic information about spoken words in memory. A central question is when such information influences linguistic processing. According to episodic models of the mental lexicon, voice-specific details influence word recognition immediately during online speech perception. Another view, the time-course hypothesis, proposes that voice-specific details influences linguistic processing only when processing is slow and effortful. The present study investigates the time-course hypothesis by employing RT-distributional analyses. We conducted a long-term repetition priming experiment using an auditory lexical-decision task. In two blocks, participants made speeded responses to existing and non-existing spoken words. In the second block, items were either repeated in the same voice, repeated in a different voice, or had not been presented in the first block. Ex-Gaussian analyses of the RT distributions in the second block revealed that voice-specific priming is reflected in distributional shifting rather than in distributional skewing. This indicates that voice-specific priming is not limited to very slow responses but that it affects fast and slow responses equally. This finding is inconsistent with a strict version of the time-course hypothesis which claims that voice-specific priming occurs only during off-line processing. Instead, voice-specific information can influence linguistic processing early on.

Prior research showed a grammatical spatial-numerical association of response codes (SNARC) effect when participants were asked to respond to singular and plural nouns. A different strand of research showed that participants in sentence completion tasks sometimes prefer to use plural verbs in conjunction with collective nouns, which are grammatically singular but denote a collection of multiple entities (e.g. cutlery). Building on these findings, we used the SNARC paradigm to investigate whether collective nouns are interpreted as conceptually singular or plural during on-line word reading. In two experiments, participants were asked to give left- and right-side responses to simple singulars, plurals and collectives in a quantity related semantic task (one vs. multiple entities?) or a verb agreement syntactic task (“is” vs. “are”?). Surprisingly, we only found grammatical SNARC effects for the more superficial syntactic task in Experiment 2, not for the deeper processing task in Experiment 1. Experiment 2 showed an interaction pattern for singulars and plurals as well as singulars and collectives with a relative left-side advantage for singulars and right-side advantage for plurals and collectives, thus indicating that reception of collective nouns automatically evokes plural quantity information, similar to grammatically plural nouns. The absence of relevant effects in Experiment 1 could be due to task ambiguity. In an effort to increase trust in the results of Experiment 2 and our interpretation thereof, we will run a replication of this study.

The question how words of multiple languages are stored in our mind has been of central research interest in the past. One prominent model addressing late bilingualism is the Revised Hierarchical Model (Kroll & Stewart, 1994) which postulates bidirectional but asymmetrical connections between separate lexical stores for L1 (native language) and L2 (second/foreign language). Additionally, these lexicons share a language-independent conceptual store, which is assumed to be involved when translating into the foreign language, but not (or at least less) when translating into the native language. This asymmetry can be, for example, observed in different translation times in dependence of translation direction (translation asymmetry) or semantic order of words (category effect). Using a sample of German native speakers with advanced English proficiency, the study strongly confirmed the model predictions regarding different processing times and preferred mental routes. In a second experiment, the same experimental paradigm was applied to a set of non-native speakers. Surprisingly, these non-native language speakers also showed the same translation asymmetry and category effect as the German native speakers, even though no native language was involved in the experiment.These findings suggest that the model not only holds true for L1 and L2, but that the mental architecture of two or more foreign languages might be similar when one of the foreign languages is currently highlighted by the experimental and general language context.

Efficient neuronal processing is essential. Thus, the Lexical Categorization Model (LCM) assumes, for the reading brain, it is beneficial to prevent semantic processing for non-words. The successful evaluations of the LCM with typical readers motivated us to implement a training study consisting of multiple sessions. The core of the LCM training is an explicit lexical decision task, i.e., is the string presented a word or not, that includes feedback on whether the response was correct. Executing lexical decisions is assumed to train the categorization process that was described by the LCM. We conducted three experiments with a three-session training for adult German language learners suffering from slow reading (N=81). Experiment 1 was the first LCM training pilot. Experiment 2 compared the training with phonics training in a randomized controlled fashion. In Experiment 3, we implemented a comparison with a variant of the LCM training (i.e., with changing fonts). We found, in general, after the LCM training, an improvement in the overall reading speed in all experiments. Also, we found an effect of word-likeness, and categorization difficulty derived from the LCM implementation. Interestingly, both effects interacted with training, e.g., the categorization effect increased with training. Finally, we found that the individual estimates of the categorization difficulty by training interaction correlated with the increase in general reading speed. This strongly suggests that we found causal evidence for the implementation of an LCM based categorization process in reading. Thus, one of the underlying processes of efficient reading is lexical categorizations.