A central issue in word naming research concerns the type of cognitive architecture that best describes the translation from orthography to phonology. This debate has centred around whether there are multiple pathways to the phonological code, for example, a lexical and sub-lexical pathway as proposed by dual-route theories, or whether only one translation pathway is required. There has been relatively little focus, however, on processes involved in translating from phonological codes to speech. A general assumption is that a speech motor program is constructed specifying the motor commands required to produce the word, and that this program is executed.

There is good reason to argue, as a number of researchers have already emphasised, that models of naming need to address the relationship between accessing phonology and executing a response more thoroughly. The involvement of response output mechanisms may be critical in understanding differences in the effects of a range of factors on naming and other tasks such as lexical decision. Output processes, for example, may contribute to effects on reaction time usually thought to reflect early stages such as stimulus identification or lexical access.

More relevant to the present concerns, however, is that there is mounting evidence against there being under all circumstances a discrete-stage partition between activating phonology and executing a response. Studies using a variety of speech production paradigms including word naming have shown that response duration is sensitive to variables that influence earlier stages of processing. Effects that permeate beyond response initiation have also been found in tasks that require manual rather than vocal responses.

These findings indicate that pre-production processes, such as those involved in translating from orthography to phonology during word naming, may not be fully resolved before initiation. Some studies, however, have failed to find significant effects on duration of factors demonstrated to influence naming reaction time.

The involvement of a sub-lexical translation pathway between orthography and phonology as proposed in dual-route theory may be critical for revealing effects on duration. Assuming a left to right conversion of graphemes to phonemes, phonemes may be passed onto production in a piecemeal fashion triggering overt vocalisation before the whole phonological code has been assembled. Some supporting evidence for this proposition already exists found that false starts during word naming sometimes included regularisations for words with irregular spelling to sound correspondences. For example, a subject when presented with CHEF may begin speaking with the regular pronunciation of the onset (i.e., as in CHEAT). This suggests that naming responses may be triggered on the basis of a sub-lexical pathway operating in a left to right fashion, while the computation of phonology is ongoing. 2 Other studies provide evidence for a left to right serial activation of phonology, although they do not directly show that execution is triggered before assembly is complete.

Frederiksen, J.R. and Kroll, J.F., 1976. Spelling and sound: approaches to the internal lexicon. Journal of Experimental Psychology: Human Perception and Performance 2, pp. 361–379. Abstract | Full Text via CrossRefFrederiksen and Kroll (1976) found that items with complex word-initial consonant clusters (e.g., CCCVC) consistently produced longer naming reaction times than items with complex word-final consonant clusters (e.g., CVCCC). Further, the greater the difference in complexity in the word-initial position, the greater the reaction time difference. This was taken to suggest that only the beginning of the letter string was processed before articulation began, revealing an effect of orthographic complexity of the initial consonant cluster on reaction time. The end of the letter string was thought to be processed for output during articulation.

The experiments described below were implemented to provide a further test of the proposition that production duration is sensitive to variables that influence pre-production processes and examine, in particular, the role of sub-lexical orthography in producing such effects. If (a) the sub-lexical pathway provides the initial part of the response before the full phonological code is available, and (b) a response criterion may be set so that the whole word need not be fully prepared prior to initiation, then the execution of word-initial segments may begin before phonological assembly is complete. If this is the case, the onset of the response will be initiated earlier in time relative to the assembly and execution of later segments in the word compared to when the sub-lexical pathway is not used. It is predicted that production duration will be longer under these circumstances given that execution needs to accommodate the increased asynchrony between when the response is triggered and when later segments become available for execution (cf. Sternberg et al., 1978). It is proposed, therefore, that subjects are able to trade-off longer response durations for faster reaction times.

Picture naming was used as a baseline condition to reveal duration lengthening of word naming responses. During picture naming, production is assumed to be based on the direct retrieval of whole word phonology from the output lexicon. There is no sub-lexical pathway by which phonology may be partially assembled and used to drive articulation before activating whole word phonology. If word naming responses are systematically lengthened as a result of sub-lexical processing, this should be revealed by comparisons with picture naming responses, other factors being equal.

Experiment 1 examines whether word naming produces longer response durations than picture naming. Intrinsic differences in duration were controlled principally by using objects names as word stimuli. Through counterbalancing materials across subjects the same responses were made in word and picture naming conditions.

Experiment 1 also included a repetition priming manipulation: some stimuli were presented for naming in an earlier part of the experimental session. Reaction times generally decrease when stimuli are repeated. This repetition priming effect may be viewed as a type of increase in the fluency of perceptual/cognitive processes underlying the response. In the context of the dual route model, priming is thought to speed lexical access, in particular, and have a minimal effect on the sub-lexical pathway. Based on this reasoning, it is expected that priming will encourage greater involvement of the lexical pathway when naming words, thereby reducing the influence of sub-lexical processing. It is predicted, therefore, that the difference in duration between words and pictures will be reduced or eliminated when naming repeated items.

The picture naming condition may be used to test whether a repetition effect on word durations might be explained by an increase in production fluency. It has been well documented that many motor skills are performed faster and more accurately with practice. If repeated responses are shorter because of production fluency, this should occur for both word and picture naming conditions.

A different-modality priming condition tested whether production fluency alone produces duration shortening. In this condition the type of stimulus (whether a word or picture), but not the response, is changed in the repeating episode. The priming effect on duration should be maintained in the different modality condition if it depends solely on reinstating the same production process regardless of access route.