Lexical access in the production of pronouns
Introduction
Speaking, most of the time, means transforming an idea or a message into a structured pattern of sounds. The speaker's goal is to produce words so that a listener can understand the message. In many theories, the transfer from ideas to words involves three main steps: message encoding, grammatical encoding, and phonological encoding (Garrett, 1975, Garrett, 1980, Garrett, 1988; Stemberger, 1985; Dell, 1986; Levelt, 1989, Levelt, 1992; Dell and O'Seaghdha, 1991, Dell and O'Seaghdha, 1992; Bock and Levelt, 1994; Bock, 1995; Levelt et al., in press). During message encoding the intended meaning of the utterance is specified prelinguistically. The message serves as input to the grammatical level of processing. During grammatical encoding, words are retrieved from the mental lexicon via a process called lexical access. As a result of lexical access, the syntactic properties of words, i.e. their lemmas, become available. The selected elements are combined into a syntactic structure. This structure is the input to the stage of phonological encoding, during which the phonological (segmental, prosodic) and phonetic shape of the utterance are generated. After the phonetic form has been computed, articulation can take place. Empirical evidence consistent with the existence of several levels of speech planning comes from speech-error data (Garrett, 1975, Garrett, 1980, Garrett, 1988; Dell, 1986; Martin et al., 1996), picture-naming studies (for a review see Schriefers et al., 1990; Glaser, 1992; Levelt et al., 1991a, Levelt et al., 1991bLevelt et al., in press; Peterson and Savoy, 1998), and electrophysiological data (Van Turennout et al., 1997, Van Turennout et al., 1998). Critical discussions of the notion of a separate syntactic (lemma) level are offered by Caramazza (1997)and Starreveld and La Heij, 1995, Starreveld and La Heij, 1996.
Most experiments on lexical access have studied the access to single nouns (e.g. Schriefers et al., 1990; Levelt et al., 1991a, Levelt et al., 1991b; Roelofs, 1992a, Roelofs, 1992b; Starreveld and La Heij, 1995, Starreveld and La Heij, 1996; O'Seaghdha and Marin, 1997; Peterson and Savoy, 1998). A few have investigated slightly more complex utterances, such as adjective noun phrases (Schriefers, 1993) and noun phrase coordinations (Meyer, 1996, Meyer, 1997; Meyer et al., 1998). In the present study, we addressed the generation of pronouns in utterances such as `The flower is red. It turns blue'. The participants in our experiments generated such sentence pairs during a picture description task. As will be explained below, our main goal was to determine whether the phonological form of the antecedent noun is activated when it is replaced by a pronoun. We will, however, first propose a working model of lexical access to pronouns.
Section snippets
Four steps from intention to pronoun access
We propose that lexical access to a pronoun involves the following four steps.
Is the antecedent phonologically activated during pronoun generation?
The aim of the experiments reported below was to investigate the phonological activation of a noun antecedent during the generation of the corresponding pronoun. In one condition, speakers produced utterance pairs such as `Die Blume ist rot. Sie wird blau.' (The flower is red. It turns blue.) In another condition, they produced pairs such as `Die Sonne ist rot. Die Blume ist blau.' (The sun is red. The flower is blue.) In order to produce the noun `Blume', its phonological representation must
The experimental paradigm
To test whether the phonological form of the antecedent noun is activated during the production of a pronoun, we used a novel version of the lexical decision during naming paradigm introduced by Levelt et al. (1991a), which is based on Stroop-like interference effects (Stroop, 1935; for a review see MacLeod, 1991). In the experiment carried out by Levelt et al., the participants' default task was to name pictures of objects. However, on some trials the naming process was disturbed by the
Participants
Thirty-two participants, aged 17–38 years, were recruited through a newspaper advertisement. They were native speakers of German. The data from three participants were excluded from further analyses because they made errors on more than 20% of the trials.
Materials
A set of unambiguous pictures was selected, such that most of the participants labeled them in the same way in the naming task. This ensured that the participants would retrieve the intended lexical item for a particular picture. In addition,
Participants
Thirty-five participants were recruited through a newspaper advertisement, and paid for participation. Their age ranged from 18 to 36 years. The data from five participants were excluded from further analysis, because they made errors on more than 20% of the trials.
Material, design, apparatus, and procedure
The same materials were used as in Experiment 1, except that in the noun condition the order of the two pictures shown on successive trials was reversed (see Appendix C). Design, apparatus, and procedure were the same as in
Conclusion
According to our working model, pronouns are used when lexical concepts are marked as `in focus' in the discourse record. Since in German, and many other languages, pronouns agree in grammatical gender with the antecedent nouns, we assume that the choice of the correct pronoun requires the corresponding lemma and the attached gender node to be selected. On the assumption that a phonological form cannot become activated in the absence of activation of the corresponding lemma, the obtained
Acknowledgements
The research reported in this paper was part of a PhD project of B. Schmitt, supported by the Max-Planck-Gesellschaft zur Förderung der Wissenschaften, München, Germany. We would especially like to thank Herbert Baumann, Inge Doehring, Gerd Klaas, John Nagengast, and Ad Verbunt for technical assistance; Jens Bölte and Pienie Zwitserlood, Department of Psychology, University of Münster, Germany, who helped us very much to carry out the experiments in German by offering one of their laboratories;
References (51)
The retrieval of phonological forms in production: Tests of predictions from a connectionist model
Journal of Memory and Language
(1988)- et al.
Stages of lexical access in language production
Cognition
(1992) Picture naming
Cognition
(1992)- et al.
Pronouns, names, and the centering of attention in discourse
Cognitive Science
(1993) - et al.
Intuitive knowledge of linguistic co-reference
Cognition
(1997) - et al.
The lexicalization process in sentence production and naming: indirect election of words
Cognition
(1983) Accessing words in speech production: stages, processes, and representations
Cognition
(1992)Lexical access in phrase and sentence production: results from picture-word interference experiments
Journal of Memory and Language
(1996)- et al.
Viewing and naming objects: eye movements during noun phrase production
Cognition
(1998) - et al.
Mediated semantic-phonological priming: calling distant relatives
Journal of Memory and Language
(1997)
A spreading-activation theory of lemma retrieval in speaking
Cognition
Exploring the time course of lexical access in language production: picture-word interference studies
Journal of Memory and Language
Syntactic processes in speech production: the retrieval of grammatical gender
Cognition
How many levels of processing are there in lexical access?
Cognitive Neuropsychology
A spreading-activation theory of retrieval in sentence production
Psychological Review
Mediated and convergent lexical priming in language production: a comment on Levelt et al. (1991)
Psychological Review
Long-term working memory. Psychological Review
Pronominalization and discourse coherence, discourse structure and pronoun interpretation
Memory and Cognition
Cited by (28)
Avoiding gender ambiguous pronouns in French
2022, CognitionCitation Excerpt :Moreover, because rose and tulipe are both feminine, both lemmas are linked to the same feminine gender node. Thus, assuming that speakers activate the lexical concept of the antecedent noun for producing a pronoun (Jescheniak et al., 2001; Schmitt et al., 1999; but see Meyer & Bock, 1999), following the activation of ROSE at the conceptual level, TULIPE receives some activation via FLOWER, and this results in lexical competition between the corresponding lemmas, rose and tulipe, at the lemma level (Levelt et al., 1999). Hence, when speakers activate rose to select the feminine gender node (F) and the pronoun lemma, elle, tulipe should interfere, and the co-activation of rose and tulipe strengthens the links between all the activated lemmas (cf. Cleland & Pickering, 2003; Pickering & Branigan, 1998), enhancing the one-to-many mapping of the pronoun.
The psycholinguistics of ellipsis
2014, LinguaCitation Excerpt :In the domain of pronoun resolution, a number of studies have tested for effects of re-access to the phonological or lexical properties of the antecedent of the pronoun at the point of pronoun resolution. For example, in a pronoun production task in German, phonological properties of the antecedent were found to be accessed when the pronoun is processed (Schmitt et al., 1999). In pronoun comprehension tasks, reading times at the pronoun are affected by lexical properties of the antecedent, such as frequency (van Gompel and Majid, 2004; Lago et al., submitted for publication).
The production of coerced expressions: Evidence from priming
2014, Journal of Memory and LanguageCitation Excerpt :Notice the “blocked” line (represented with “X”) from the lemma for drink to the coerced constituent-structure representation, which relates to the observation in Experiment 4 that coercion priming was enhanced despite the absence of an overtly-expressed coerced verb. Our assumption that the lemma of the coerced verb is retrieved from the lexicon even though it is not ultimately expressed is compatible with evidence for lexical retrieval of unexpressed elements, for example pronoun production, in which both the lemma and phonological form for a full noun appear to be activated even when its referent is realized as a pronoun (Schmitt, Meyer, & Levelt, 1999). Our proposal that semantic-level processes underlie the enhanced priming in the absence of (overt) lexical repetition is compatible with the finding of a translation-equivalent boost to priming (Bernolet, Hartsuiker, & Pickering, 2012; Cai, Pickering, Yan, & Branigan, 2011; Schoonbaert, Hartsuiker, & Pickering, 2007), whereby enhanced priming occurs on the basis of shared semantic representations even in the absence of shared lemma representations, as well as the semantic boost to priming identified by Cleland and Pickering (2003).
A processing-centered look at the contribution of givenness to durational reduction
2012, Journal of Memory and LanguageCitation Excerpt :Crucially, the trigger approach posits a categorical representation of information status, and that the model selects a form or rule based on the information status category. Although the Schmitt et al. (1999) model includes only a binary contrast between “focus” and “not focus,” it could in principle include finer-grained information status differences. For example, Levelt (1989) proposes that speakers model the information status of referents with four categories: (a) inaccessible, (b) accessible, (c) in discourse model, and (d) in focus.
Phonological activation of ignored pictures: Further evidence for a cascade model of lexical access
2005, Journal of Memory and Language