Kathy Conklin

Role(s): Lecturer, Academic

Current research

Although my research reflects a diverse set of interests the focus is on lexical-semantic representation and its interface with syntax and sentence or discourse processing. Much of my research also addresses how representation and processing in one language is influenced by knowledge of another language. My research has implications for models of both monolingual and bilingual word recognition, models of sentence processing, and syntactic theory. Below I discuss four current and ongoing research investigations: (1) grammatical gender and anaphor resolution, (2) pronoun resolution in Chinese, (3) interlingual homographs and homophones, and (4) multi-word units: idioms and binomials. 

Grammatical gender and anaphor resolution

A number of cues have been implicated in pronoun resolution. For example, a pronoun’s morphosyntactic features (person, number, gender) indicate its potential referents in a discourse (Chomsky, 1981). When only one referent from the discourse matches in morphosyntactic features, pronoun resolution is speeded (Erlich, 1980). When the referent of a pronoun cannot be determined on the basis of morphosyntactic features, other cues such as prominence and structural parallelism are used. A pronoun that refers to the most prominent member of the current focus of attention is easier to interpret (Gordon & Chan, 1995), as is one in a parallel syntactic position (Chambers & Smyth, 1998).

In English, the pronouns he and she refer to entities that have natural gender (i.e. people). Inanimate objects are referred to with the pronoun it. In Dutch, which uses grammatical gender, not only people but also inanimate objects (tractor or vlieger ‘kite’) may be referred to using the masculine pronoun hij ‘he’. Eyetracking studies with high proficiency Dutch-English bilinguals and English monolingual control participants were conducted to determine whether Dutch gender information influences English pronoun resolution. Participants listened to an English discourse while viewing a corresponding visual scene containing a cartoon character (Donald Duck) and an inanimate object (tractor). Interpretation of pronouns was investigated following auditory presentation of inanimate nouns and the subsequent pronoun he (The tractor will be driven by Donald Duck. He is in the other field.). Bilinguals had increased looks and fixation durations to inanimate objects for cognates (tractor) for 2000 ms following pronoun onset. There were only increased looks to inanimate noncognates (kite/vlieger) for 400 ms. Monolinguals only looked at the animate character. These results indicate that Dutch-English bilinguals activate information about Dutch gender when processing spoken English discourse.  

Dutch speakers are becoming increasingly unaware of grammatical gender. Therefore, some of my ongoing research further explores the finding with Dutch-English bilinguals in other languages, namely French and German. This will increase the likelihood of finding an effect for non-cognates. This research also explores processing by participants at different levels of expertise in order to ascertain to what extent proficiency attenuates the effect. The paradigm is also being used to explore the age at which German monolinguals use cues such as structural parallelism, and natural and grammatical gender in anaphor resolution.  

Pronoun resolution in Chinese

The above work on anaphor resolution has led to a more general interest in pronoun processing. To date most investigations on anaphor resolution have been done on English and other Indo-European languages. Recent work indicates that despite linguistic differences between English and Mandarin, structural parallelism and morphosyntactic cues are used in similar ways in the two languages. However, Mandarin also makes use of zero pronouns. Yang et al. (1999) investigated the role of structural parallelism by examining processing of subject pronouns that referred to first or second mention entities in a previous sentence when the referents were unambiguous (had different genders) or ambiguous (had the same gender) and when pronouns were overt or zero. Using a measure of whole sentence reading time, they found that in ambiguous sentences there was no effect of zero vs. overt pronouns in either the first or second mention conditions. My current work further explores these basic findings using a region-by-region moving window paradigm to give a more fine-grained measure of the processing effort. Similar to Yang et al. and the literature on English, results indicate that overt pronouns are processed quickly in both first and second mention conditions in unambiguous contexts that provide a gender agreement cue. However, processing is significantly slower in ambiguous contexts when referring to a second rather than a first mentioned entity. In contrast to Yang et al.'s findings, my results indicate that when referring to a first mentioned entity with zero pronouns initially processing is slowed, but that by the end of the sentence there is no significant processing cost. When referring to a second mentioned entity, contexts with zero pronouns require significantly more processing time throughout the sentence. These findings indicate that in ambiguous sentences, in both the first and second mention conditions, overt pronouns provide a better cue than zero pronouns for determining an antecedent.

Interlingual homographs and homophones

Previous research on word recognition has shown that knowledge of our L1 often interacts with processing of our L2, regardless of our level of proficiency (e.g. Dijkstra and van Heuven, 2002). Despite the theoretical importance of this interaction, most research to date has explored this issue by investigating individual word processing. My research investigates how word frequency, language proficiency, and two types of context (linguistic and sentence) affect the activation of multiple meanings of interlingual homographs (coin English ‘money’, French ‘corner’) in sentence contexts. Results from native French speakers with a high English proficiency showed that when only the language of processing (linguistic context) constrains the interpretation of an interlingual homograph (While walking the little boy found a coin.) both the language appropriate and inappropriate meanings are activated. When the sentence context constrains the interpretation of the interlingual homograph (The thing with the lowest monetary value is a coin.) only the contextually appropriate meaning is activated. This pattern of results is consistent with the predictions of a proposed bilingual variant of the Reordered Access Model (RAM) (Duffy, et al., 1988). When processing in L2, the activation of L2 meanings of interlingual homographs is boosted sufficiently for them to become competitors with the normally stronger L1 meanings. As a consequence, it takes longer for participants to select the appropriate L2 meaning. When the sentential context strongly supports the L2 meaning, the L2 meaning of the interlingual homograph behaves like the more frequent, dominant meaning of an ambiguous word in monolingual studies, and is selected before the other meaning can effectively compete with it. An assumption of the bilingual variant of the RAM is that there is an asymmetry in the resting activation levels of L1 and L2 meanings of interlingual homographs. A prediction that stems from this assumption is that there should be no bilingual subordinate bias effect when nonnative speakers process interlingual homographs in an L1 sentential context; however, recent processing in one’s L2 could briefly raise the activation of L2 meanings such that they might compete with L1 meanings. Although overall, I did not observe a bilingual subordinate bias effect when nonnatives processed in their L1, split half analyses of experimental results revealed that recent L2 processing briefly boosts the activation of L2 when processing in L1. Put in terms of the debate on exhaustive versus selective access, these findings show that processing in one’s L1 is sufficient to boost activation of lexical items such that it lowers the competitiveness of L2 meanings so that lexical access in L1 looks selective.  

Ongoing research investigates effects of cross-script interactions in Chinese and in Chinese-English speakers. Chinese words can be written using two distinct scripts, namely Chinese characters and Hanyu Pinyin. Hanyu Pinyin is the romanization system for Standard Mandarin that is used to teach characters in schools in Mainland China. Chinese characters and pinyin are presented in textbooks until children are about 10 years old. Children and adults continue to use pinyin in everyday life as it is used to enter characters on computers, text messages, etc. To date little research has been conducted on the role of pinyin in Chinese reading and in reading English as a second language. This research investigates how and when pinyin influences reading. Current researcher investigates the processing of Chinese-English interlingual homographs and homophones (FAN represents the Chinese character meaning "sail") with naming tasks, ERPs, and eyetracking. By measuring response latencies, brain responses (ERPs), and eye movements this research will provide a comprehensive picture of Chinese lexical representation and visual word recognition and will have important implications for both monolingual and bilingual models of visual word recognition and sentence processing. 

 Mulit-word units: idioms

It is generally accepted that we store representations of words in our mental lexicon, however, what exactly is stored remains an open question. For example, do we store the word dog as well as its plural form dogs, or do we only store dog and have a rule (NOUN + s = plural) to compute the plural form? A similar question arises regarding the storage vs. computation of multi-word units, wherein a single meaning is attached to a string of words (e.g., phrasal verbs, compounds, and idioms). There are compelling reasons for proposing that the brain stores and processes these multi-word units as wholes, as vast amounts of knowledge can be stored in long-term memory, but only relatively small amounts can be processed in real-time. In effect, the brain may make use of a relatively abundant resource (long-term memory) to compensate for a relative lack in another (working-memory) by storing a number of frequently used multi-word units as wholes. These can be easily retrieved and used without the need to compose them on-line through word selection and grammatical sequencing. This means there is less demand on cognitive capacity because the multi-word units are ‘ready to go’ and require little or no additional cognitive processing. While this may be the case when processing a first language, a further question is whether processing a second language works in the same way.

Current investigations look at the processing of idioms by native and nonnative speakers. Thus far I have examined processing of idioms used figuratively (ring a bell, ‘remind’) and literally (ring a bell, ‘produce sound with metal object’), as well as matched novel phrases (ring the bell) in disambiguating contexts. Consistent with previous research, findings indicate that native speakers process idioms faster than novel language. Interestingly, idioms’ figurative and literal meanings were processed with the same speed and this held true before and after the recognition point. Although context facilitated the processing of the low frequency literal meaning, further study is needed to determine whether figurative and literal meanings show similar processing patterns when context does not provide a disambiguating cue and whether idioms are stored holistically or computed on-line. For nonnative speakers, longer fixation durations and more fixations were found for idioms used figuratively than literally, suggesting that the link between the idiom and its figurative meaning (ring a bell = remind) is not as strong as the link between the forms and meanings of the individual lexical items of the idiom (ring + a + bell = produce sound). Further, it appears that nonnatives cannot use context effectively to activate the appropriate, figurative meaning, suggesting that even in the presence of strong contextual cues, the figurative meaning does not compete for selection with the literal one.

View my video 'The Eyes Trackers' on YouTube or my vodcast about '2nd Language Processing'

Expertise summary