| Sentence production requires the translation of a non-linguistic message into an ordered sequence of words. This process includes a two-level grammatical encoding stage, consisting of a functional level and a positional level.;Experiments 1--3 investigated role and position assignment by asking if the structural similarity of constituents affects functional- and positional-level processing, measured as effects on rates of exchange errors: production errors in which constituents swap roles or positions (e.g., She donated a library to the book; intended: She donated a book to the library). Phrase exchanges, which arise at the functional level, and word exchanges, which arise at the functional or positional level, were analyzed. Garrett (1975, 1980} found that exchanging words and phrases are almost exclusively similar in grammatical class and in their relative role or position in an utterance. This suggests that if other, more complex types of grammatical similarity not directly tested by Garrett are experimentally manipulated, they will affect exchange error rates. Bock and Levelt's (1994) model further specifies this hypothesis: If grammatical encoding establishes the roles of, relationships between, and positions of constituents, then phrases should be more likely to exchange if they have the same rather than different grammatical structures; and word exchanges should be more likely for words in similarly- rather than differently-structured phrases.;In Experiments 1--3, participants described drawings of common objects (e.g., an apple with a spot on it), presented in different color schemes, with two noun phrases (NPs) linked by a preposition. NPs in Experiment 1's and Experiment 2's homogeneous condition (e.g., the brown spot on the blue apple) had the same structure: They contained the same grammatical classes of words (e.g., a determiner, an adjective, and a noun) in the same relationships. Homogeneous-condition NPs also had the same semantic content, in that their words described the same abstract kinds of things: an object and a color, or just an object. NPs in the heterogeneous condition (e.g., the brown spot on the apple) differed in structure and semantic content. In Experiment 3, responses varied only in structural similarity. Homogeneous NPs (e.g., the brown spot on the blue apple) had the same structure and type of semantic content, while heterogeneous NPs (e.g., the brown spot on the apple that's blue) had the same type of semantic content, but different structures.;In Experiment 1, phrase and word exchanges showed similarity effects: The likelihood of an exchange error was greater for homogeneous responses. Experiment 2 determined that Experiment 1's similarity effects were not due to its color scheme. In Experiment 3, there were no similarity effects, suggesting that Experiment 1's and Experiment 2's similarity effects could not be due to structural similarity alone. Semantic content similarity or semantic content in conjunction with structural similarity was necessary to affect exchange errors during grammatical encoding.;Experiments 4--6 investigated content word selection in determiner-adjective-noun NPs.;In Experiment 4, participants described drawings of common objects (clothing and shapes), each with a color or pattern applied (e.g., the purple dress, the striped triangle), in a picture-word interference paradigm with semantically related, phonologically related, and unrelated written distractors presented at six stimulus onset asynchronies (SOAs). Two distractor effects were analyzed: semantic interference and phonological facilitation. Semantic interference was measured as the difference in production latency between the semantically related and unrelated conditions. Significant semantic interference indicated that the semantically related distractor had affected lemma selection. Phonological facilitation was measured as the latency difference between the phonologically related and unrelated conditions. Significant phonological facilitation indicated that the phonologically related distractor had affected lexeme selection.;There was semantic interference at negative SOAs (when the distractor appeared before the drawing), marginal for nouns and significant for adjectives. It began at a more negative SOA for adjectives, suggesting that for lemmas, adjective selection began first and overlapped with noun selection. So, the lemmas were selected in the order that they were ultimately produced. This selection order may be driven by the knowledge that adjectives are canonically produced before nouns in English, or by attention directed first to the picture's attribute. More research is necessary to investigate these explanations. There was no adjective phonological facilitation, but marginal noun facilitation at positive SOAs (when the distractor appeared after the drawing) suggested that noun lexeme selection occurred after noun lemma selection, replicating past research (Schriefers, Meyer, & Levelt, 1990).;Experiments 5 and 6 were control experiments using the same stimuli but single-word responses: isolated nouns in Experiment 5 and isolated adjectives in Experiment 6. Consistent with past research (Schriefers et al., 1990), Experiment 5 found early noun lemma selection. Experiment 6 found early adjective lemma selection, followed by adjective lexeme selection. (Abstract shortened by UMI.). |
