RUNNING HEAD: Strategic Control in Naming Coarse Strategic Control in Word Reading: Evidence from Tempo Naming and a Computational Model

We investigate the hypothesis that lexical and rule knowledge can be strategically emphasized in word naming based on task demands, in contrast with the hypothesis that a time criterion to respond can be strategically manipulated. We devised the tempo naming to investigate the extent to which readers have precise control over response initiation in naming. Relative to baseline performance in the standard naming task, subjects were induced to respond with faster latencies, shorter durations, and lower levels of accuracy by instructing them to time response initiation with an experimentally controlled tempo. The effects of printed frequency and spelling-to-sound consistency were attenuated for latencies in the tempo naming task, and the number of regularization errors remained constant with faster tempos, while the frequency of word, articulatory, and other errors increased. We interpret the pattern of effects as inconsistent with existing models of word reading. We propose that cognitively coarse parameters serve as the levers of strategic control over response initiation in word reading. We account for the tempo naming results in a connectionist model that embodies coarse strategic control by scaling the net input to all units the network via input gain, and we extend the model to account for a set of stimulus blocking effects previously interpreted as evidence for a time criterion. In the general discussion, we address how coarse strategic control and its current implementation relate to a broader set of issues in word reading and other domains of strategic control. Strategic Control in Naming 79 Interest in strategic control over processing in word reading has grown in recent years (Baluch & Besner, 1991; Colombo & Tabossi, 1992; Jared, 1997; Lupker, Brown, & Colombo, 1997a; Monsell, Patterson, Graham, Hughes, & Milroy, 1989; Paap & Noel, 1991). Research on the empirical and theoretical aspects of strategic control has addressed questions concerning the phenomena itself, as well as the architecture and processes of word reading more broadly construed. One contributing factor to the popularity this research has been its relevance to dual-process theories of the mapping from orthography to phonology (e.g., Coltheart, Curtis, Atkins, & Haller, 1993; Plaut, McClelland, Seidenberg, & Patterson, 1997). The proposal of two distinct routes for generating the pronunciation of a letter string has prompted researchers to suggest that subjects can selectively emphasize or de-emphasize one or both routes depending on task demands. A number of studies have shown effects that have been explained as a de-emphasis of rule knowledge when that knowledge interferes with performance (Coltheart & Rastle, 1994; Herdman, 1992; Monsell et al., 1989), and likewise, a de-emphasis of lexical knowledge when it could interfere (albeit the latter has been in languages with transparent orthographies relative to English; Tabossi & Laghi, 1992; Baluch & Besner, 1991; Colombo & Tabossi, 1992). These studies have been used as evidence against a single-route account of word reading. Recently, researchers have provided an alternative to the route emphasis account (Jared, 1997; Lupker et al., 1997a). Rather than manipulate route processing, subjects are hypothesized to strategically control a time criterion to initiate the naming response. Lupker et al. and Jared argued that manipulation of a time criterion can account for a number of blocking effects that had previously been used to argue for the route emphasis account. The time criterion hypothesis, as they proposed it, was problematic for two reasons. First, manipulation of a time criterion did not explain at least one aspect of the results from the Lupker et al. (1997a) study. Second, the authors did not present an explicit mechanism of how a time criterion is applied within and across trials, and it is unclear what version would best account for the relevant phenomena. In the current study, we investigate the hypothesis of strategic control over a time criterion by introducing a novel methodology, called tempo naming, designed to manipulate explicitly the initiation of a naming response. The results from three tempo naming experiments lead us to argue for an alternative to the time criterion and route emphasis accounts of strategic blocking effects. We propose that strategic control over response initiation is accomplished by changing the rate of information integration across the central processes involved in word reading. We embody this concept in a connectionist framework by controlling a scale parameter on the net input to all units in the network, termed gain (Cohen & Servan-Schreiber, 1992). We argue that results from the tempo naming experiments, as well as Lupker et al.’s (1997a) stimulus blocking experiments, are best accounted for by changes in a single parameter (level of gain) that alters processing across multiple levels of representation. We show that the manipulation of gain, when applied uniformly to all processing units, allows for complex, qualitative changes in behavior due to the non-linear interactions between and within levels of representation. We present a distributed, connectionist model of word reading that instantiates the gain hypothesis and accounts for the main results of the tempo naming experiments, as well as the main results from the stimulus blocking study by Lupker and his colleagues. In the General Discussion, we broaden the scope of our work in the context of considering a wider range of phenomena in strategic control and word reading. Empirical and Theoretical Background It takes roughly 400--600 ms for a skilled reader to begin the pronunciation of a single, clearly printed word. This ballpark range comes from a long history of speeded word naming studies in which subjects were asked to pronounce a printed word under instructions to the effect of “as quickly and accurately as possible.” The speeded word naming task has been used to examine a wide variety of theoretical issues such as the following: processes that map orthography to phonology (Glushko, 1979; Seidenberg, Waters, Barnes, & Tanenhaus, 1984), organization of the lexicon (Forster & Chambers, 1973; Frederiksen & Kroll, 1976), semantic, phonological, and orthographic priming (Forster & Davis, 1991; Tabossi & Laghi, 1992; Taraban & McClelland, Strategic Control in Naming 8