Masked form priming in writing words from pictures: Evidence for direct retrieval of orthographic codes
Introduction
While substantial research has been devoted to lexical access in spoken picture naming, few studies have addressed the corresponding problem in written picture naming. It is fair to say that the study of speech production is more advanced than the study of written production. This “dearth of experimental research” for low level processing components in writing can be explained by the fact that current studies have focused on higher processing levels, such as planning (Hayes and Flower, 1980), or revising (Flower et al., 1986), and that the investigation of writing had to cope with serious methodological problems (see Fayol, 1997, for a review).
In the present study, we investigated lexical access in written picture naming. We assumed that current views of speech production provide a general theoretical framework from which hypotheses specific to writing can be derived. It is generally held that speech production (e.g., naming a word from a picture) involves several processes (see Bock and Levelt, 1994, for a review). After selection of the concept, activation spreads to semantic features and to the appropriate lemma nodes Levelt et al., 1991a, Schriefers et al., 1990. In turn, the lemma node activates its corresponding phonological representation (i.e., a lexeme). The final step consists of the computation of, or access to, the articulatory gestures Levelt, 1989, Levelt and Wheeldon, 1994. Although some controversies exist about the relative time course of the processing levels (e.g., Dell and O’ Seaghdha, 1991, Humphreys et al., 1988, Levelt et al., 1991b), this modeling of speech production is supported by various findings, including the analyses of speech errors in normal subjects and brain-damaged patients Dell and Reich, 1981, Fay and Cutler, 1977, Fromkin, 1971, Garrett, 1975, Garrett, 1980, Garrett, 1982, Henaff-Gonon et al., 1989, Kay and Ellis, 1987, the tip-of-the-tongue phenomena Brown and McNeill, 1966, Jones and Langford, 1987, as well as by several experimental studies involving normal subjects Levelt et al., 1991a, Schriefers et al., 1990.
It is generally assumed that writing and speech production systems share the conceptual-semantic level Bonin et al., in press, Caramazza and Hillis, 1990, Hillis and Caramazza, 1991, Hillis and Caramazza, 1995, Rapp and Caramazza, 1994. Written production would subsequently involve a graphemic level Ellis, 1982, Margolin, 1984 specifying the syllable structure and the identity of individual graphemes (Caramazza and Miceli, 1990). Henceforth, this processing level will be referred to as the orthographic lexeme level. Letters would then be specified in terms of spatial description at the allographic levelEllis, 1982, Weekes, 1994, and the graphic motor patterns would be retrieved at the graphic level(Van Galen, 1980). Such distinctions between the processing levels are supported by analyses of errors in normal and brain-damaged patients (Baxter and Warrington, 1986, Ellis, 1979, Goodman and Caramazza, 1986, Miceli et al., 1985 for a review, Bonin, 1997).
A current debate concerning the involvement of lexical access in writing is related to the role played by phonological information. Traditionally, it is assumed that written language skills rely on spoken language knowledge and processes Geschwind, 1969, Luria, 1970. Hence, access to orthography would be dependent on the prior retrieval of the lexical phonological representation of the word. This hypothesis is referred to as the obligatory phonological mediation hypothesisRapp and Caramazza, 1994, Rapp and Caramazza, 1997, Rapp et al., 1997. This phonological mediation hypothesis mirrors the fact that spoken language precedes, ontogenetically and phylogenetically, written language (Scinto, 1986). In addition, this hypothesis is seen as consistent with the observations of phonologically based errors in written productions such as homophone substitutions (e.g., the word “there” written “their”; the word “often” written “oven”; Aitchison and Todd, 1982) and with our introspective experiences of inner speech that accompanies writing (Hotopf, 1980).
In contrast, according to the orthographic autonomy hypothesisRapp and Caramazza, 1994, Rapp and Caramazza, 1997, Rapp et al., 1997, orthographic and phonological representations can be addressed independently in language production. Hence, contrary to the obligatory phonological mediation hypothesis, retrieval of the orthographic codes does not require access to phonology.
The obligatory phonological mediation hypothesis is faced with various problems such as the production of the correct orthographic form of homophones (i.e., seen vs. scene, Largy et al., 1996), silent graphemes (i.e., h in the French word harpe), and double letters (i.e., pp in the French word nappe). Moreover, neuropsychological reports indicate that written performance can be relatively spared when compared to spoken production Assal et al., 1981, Hier and Mohr, 1977, Lhermitte and Derouesné, 1974, Patterson and Marcel, 1977, Rapp and Caramazza, 1994. For example, Lhermitte and Derouesné (1974) described a patient who was 74% correct in written production but only 8% correct in spoken production. Such cases are problematic for the phonological mediation hypothesis because it is difficult to argue that spoken neologistic responses form the basis for the retrieval of correct written responses. Moreover, some of the individuals producing neologisms have little or no trouble in reading aloud (Miceli and Caramazza, 1993), therefore ruling out the hypothesis that neologisms arise at an output stage. Additional difficulties for the obligatory phonological mediation hypothesis come from neuropsychological observations that some patients produced semantic errors in reading and spoken picture naming but not in writing (Caramazza and Hillis, 1990). These observations are compatible with the view that semantic errors in spoken production lie at the level of phonological encoding, and that orthographic lexemes can be accessed without phonological mediation. Indeed, in the phonological mediation hypothesis, semantic errors should also have been observed for written production.
Although the orthographic autonomy hypothesis seems suitable to account for neuropsychological data, we are not aware of clear confirmative evidence from normal subjects. In the present study we examined whether, in normal subjects, phonology is obligatorily accessed when writing picture names. The masked form priming paradigm allows us to evaluate, directly, the orthographic autonomy hypothesis while, at the same time, avoiding the use of predictive strategies. In this technique, initially developed to investigate visual word recognition Evett and Humphreys, 1981, Forster and Davis, 1984, the prime visibility is reduced by using short prime duration and forward and backward masking. In such priming conditions, the prime is generally not available for conscious report, and it is therefore unlikely that the participants use the primes strategically to predict the targets (e.g., to expect the presentation of a picture depicting a “rose” upon the presentation of the nonword prime “roze”). This paradigm was recently used by Ferrand et al. (1994) to study spoken picture naming and word naming.
In Ferrand et al.'s study (Ferrand et al., 1994), three different nonword primes were used. In a first condition (pseudohomophone prime condition), the primes were homophonic with the picture names, and shared most of their letters with the picture names. In a second condition (orthographic prime condition), primes were also orthographically related to the picture names, but were not homophonic although phonologically similar with the picture names. Finally, in a third – control (control prime condition) – the orthographic and phonological overlap between primes and picture names was restricted to the first consonant or consonant cluster. The results showed that spoken picture naming was facilitated by pseudohomophone primes when compared to orthographic primes and to controls. The latter two conditions gave rise to similar performances. Thus, spoken picture naming was facilitated by the preactivation of phonological representations, but not by the preactivation of orthographic information. As the priming effect for spoken picture naming was similar in size when the prime was either the picture label or when it was a pseudohomophone, Ferrand et al. (1994) concluded that the pseudohomophone priming effect resulted from the preactivation in memory of the phonological representation corresponding to the picture name.
The aim of the present study is to investigate the role of phonological and orthographic codes in written picture naming through the use of the three priming conditions employed by Ferrand et al. (1994) in spoken picture naming. Hence, pictures were primed by pseudohomophones and orthographically related nonwords (referred to as “pseudohomophone primes”), orthographically related nonwords but less phonologically related nonwords (referred to as “orthographic primes”), or nonwords that were orthographically and phonologically related to the target picture names only on the first consonant or consonant cluster (referred to as “control primes”).
Fig. 1 presents a general working model of lexical access in written picture naming that clarifies the predictions examined in the experiments. According to this model, the presentation of the picture causes activation of structural representations in memory (Humphreys et al., 1995). Then, activation flows from structural representations to semantic representations. According to the hypothesis of the direct retrieval of orthographic information, activation would spread directly from semantic features to orthographic codes, and subsequently to the corresponding allographic codes. The optional phonological route to orthography is represented in Fig. 1 by the arrow between phonological and orthographic codes (via lexical or sublexical links).
Supposing that orthographic codes are activated through phonology, written picture naming should be facilitated when the nonword primes are homophonic with the picture label. Indeed, because the pseudohomophone primes have a larger phonological overlap with the picture names than the orthographic primes, they should facilitate more the retrieval of the phonological codes, and the subsequent access to the associated orthographic specifications. On the contrary, if, as depicted in Fig. 1, orthographic codes can be accessed directly from the picture concept, then priming with orthographic primes should be as efficient as priming with pseudohomophone primes, as these two primes are equally similar, orthographically, to the picture names. A preliminary experiment showed that a prime duration of 17 ms (one refresh cycle) was too short to generate reliable priming effects. Therefore, a prime duration of 34 ms (two refresh cycles) was used in Experiment 1. To permit direct comparisons with the data described by Ferrand et al. (1994) in spoken picture naming, all nonword primes and picture labels were identical to those used by Ferrand et al. (1994).
Section snippets
Experiment 1
In Experiment 1, target pictures were primed for 34 ms with the three different types of nonwords previously described.
Experiment 2 – Priming with nonwords using a different pattern mask (34 ms)
The purpose of Experiment 2 was to replicate Experiment 1 using a pattern mask that maximally prevents prime identification.
Experiment 3 – Priming with nonwords (51 ms)
Experiments 1 and 2 suggest that orthographic information in written picture naming can be accessed without phonological mediation. Whereas pseudohomophone primes improved spoken picture naming when presented for 29 ms in Ferrand et al.'s (1994) study, Experiments 1 and 2 showed that homophony between primes and picture names did not lead to an additional advantage in written picture naming. However, although the data reported by Ferrand et al. (1994) suggest that a 34 ms prime duration is
General discussion
The present study was undertaken to investigate the role of orthographic and phonological information in written picture naming. As we were primarily interested in fast and obligatory processes involved in written picture naming, the masked form priming paradigm was thought to be appropriate as the brief prime exposure durations generally prevent conscious identification.
Three experiments were designed to determine the nature of the information involved in writing words from pictures. Two
Acknowledgements
The authors wish to thank Brenda Rapp, Gezinus Wolters, and three anonymous reviewers for helpful comments on previous versions of the manuscript, Magali Nieddu for running Experiment 1, Annie Zielinski for running Experiment 2, Sandrine Brunie for running Experiment 3 and Fabienne Baudot for drawing the model depicted in Fig. 1.
References (62)
- et al.
Dissociations in aphasia: A case report
Brain and Language
(1981) - et al.
The “tip of the tongue” phenomenon
Journal of Verbal Learning and Verbal Behavior
(1966) - et al.
Where do semantic errors come from?
Cortex
(1990) - et al.
The structure of graphemic representations
Cognition
(1990) The language-as-fixed-effect fallacy: A critique of language statistics in psychological research
Journal of Verbal Learning and Verbal Behavior
(1973)- et al.
Stages in sentences production: An analysis of speech error data
Journal of Verbal Learning and Verbal Behavior
(1981) - et al.
Incongruous oral and written naming
Brain and Language
(1977) - et al.
Mecanisms for accessing lexical representations for output: Evidence from a category-specific semantic deficit
Brain and Language
(1991) - et al.
Do speakers have access to a mental syllabary?
Cognition
(1994) - et al.
Cognitive analysis of a case of pure dysgraphia
Brain and Language
(1985)
The modality-specific organization of grammatical categories: Evidence from impaired spoken and written sentence production
Brain and Language
Exploring the time course of lexical access in langage production: Picture-word interference studies
Journal of Memory and Language
Ideational agraphia a single case study
Journal of Neurology, Neurosurgery and Psychology
A rose is a REEZ: The two-cycles model of phonology assembly in reading English
Psychological Review
Produire des mots isolés oralement et par écrit
Revue de Neuropsychologie
Role of phonological and orthographic codes in picture naming and writing: An interference paradigm study
Current Psychology of Cognition
PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using Macintosh computers
Behavior Research Methods, Instruments, and Computers
Mediated and convergent lexical priming in language production: A comment on Levelt et al
Psychological Review
Slips of the pen
Visible Language
The use of abstract graphemic information in lexical access
Quarterly Journal of Experimental Psychology
Malaproprisms and the structure of the mental lexicon
Linguistic Inquiry
A study of masked form priming in picture and word naming
Memory and Cognition
Amorçage phonologique masqué et dénomination
L'Année Psychologique
Detection, diagnosis, and the strategies of revision
College Composition and Communication
Repetition priming and frequency attenuation in lexical access
Journal of Experimental Psychology: Learning, Memory, and Cognition
Cited by (44)
The time course of lexical and sublexical phonological activation in Chinese written production
2022, Biological PsychologyCitation Excerpt :Some other agraphic patients were reported to produce different spoken and written names in response to the same picture (Alario et al., 2003; Caramazza, 1997; Miceli et al., 1997), e.g., the patient would say “pincers” but write “saw” when presented with a picture of a pliers. Furthermore, a few behavioral studies failed to find phonological effect (Bonin et al., 1998; Shen et al., 2013) or found earlier orthographic than phonological activation (Zhang & Wang, 2015, 2016) in written production. These findings suggest that orthographic activation is not dependent on phonological activation, supporting the orthographic autonomy hypothesis.
Masking procedures can influence priming effects besides their effects on conscious perception
2019, Consciousness and CognitionCitation Excerpt :Masked priming examines whether the operation of a specific information-processing system relies on conscious awareness of the processed stimuli and provides a tool to determine the role of consciousness for information processing (Dehaene, Changeux, Naccache, Sackur, & Sergent, 2006). With this approach a number of studies examined the language processing system with semantic (e.g., Dell’Acqua & Grainger, 1999), phonological (e.g., Rastle & Brysbaert, 2006), and orthographical priming (e.g., Bonin, Fayol, & Peereman, 1998). Part of these studies provided evidence for the view that semantic processing of words improves with the visibility of the effective stimuli.
Word writing in Spanish-speaking children: Central and peripheral processes
2018, Acta PsychologicaCitation Excerpt :Initially, it was hypothesised that the central (spelling) and peripheral (handwriting) processes were independent, so the onset of the motor output was an indicator that the central processes has been completed, regardless of the route followed to access the orthographic representation (Ellis, 1988; Margolin, 1984). From this hypothesis, two lines of study emerged: research into the central orthographic processes through measures such as writing latencies (Bonin, Fayol, & Peereman, 1998; Bonin & Meot, 2002), and research about peripheral orthographic processes through kinematic measures such as writing durations—including different segments, as word, letter or stroke (Meulenbroek & Van Galen, 1988; Portier, Van Galen, & Meulenbroek, 1990). Written latency (WL) is considered to be the time between presentation of the stimulus (auditory in spelling-to-dictation or visual in copying) and first contact of the pen with the paper.
Tracking the time course of lexical access in orthographic production: An event-related potential study of word frequency effects in written picture naming
2016, Brain and LanguageCitation Excerpt :In Baus et al. (2013) participants typed the names of pictures into a keyboard located on their lap, and response latencies (time interval between stimulus presentation and first key stroke) were relatively slow, with a mean of 1441 ms (the only study that we are aware of which has used a picture naming task with typed responses reported by Pinet et al., 2015, showed average latencies of 1084 ms). By contrast, numerous studies on handwritten word production (e.g., Bonin, Fayol, & Peereman, 1998; Perret & Laganaro, 2013; Qu, Damian, Zhang, & Zhu, 2011) which have resulted in latencies (interval from stimulus presentation to first contact of pen with writing tablet) similar to those found in spoken picture naming (i.e., 800–1000 ms). Indeed, Perret and Laganaro (2013) recently compared latencies of spoken and written responses directly, and found that written responses were somewhat (∼100 ms) slower when writers could monitor their written responses.
Abstract graphemic representations support preparation of handwritten responses
2013, Journal of Memory and LanguageCitation Excerpt :The existence of a direct route from semantics to orthography, as stipulated in the orthographic autonomy account, does not rule out potential influences of phonology, however: access to orthographic codes in writing could be supported both by a direct link from semantics to handwritten spelling, and an indirect link from semantics to phonology to handwritten spelling. Although direct empirical evidence is still scant, an increasing number of studies have documented phonological constraints on orthographic output tasks (but see Bonin, Fayol, and Peereman (1998) for a failure to obtain phonologically based masked priming effects in written picture naming). For instance, Bonin, Peereman, and Fayol (2001) manipulated the consistency of phonology–orthography mappings in picture names in order to identify to contribution and locus of phonological variables in written picture naming.
Does brand spelling influence memory? The case of auditorily presented brand names
2013, Journal of Consumer PsychologyCitation Excerpt :Conversely, some phonemes (e.g., /_@g/ as in tag) are represented by only one grapheme (ag), so they have consistent mappings, which are used in different words (e.g., tag, rag). Consistency effects have been observed in faster and more accurate responses for consistent than for inconsistent words in a number of different tasks, such as reading and spelling performance (Coltheart & Leahy, 1992; Content, 1991; Ziegler et al., 1997) and writing to dictation (Bonin, Fayol, & Peereman, 1998; Bonin, Peereman, & Fayol, 2001). The number of possible sound-to-spelling mappings determines the process by which sounds are more reliably spelled.