Modeling lexical decision: the form of frequency and diversity effects

What is the root cause of word frequency effects on lexical decision times? W. S. Murray and K. I. Forster (2004) argued that such effects are linear in rank frequency, consistent with a serial search model of lexical access. In this article, the authors (a) describe a method of testing models of such effects that takes into account the possibility of parametric overfitting; (b) illustrate the effect of corpus choice on estimates of rank frequency; (c) give derivations of nine functional forms as predictions of models of lexical decision; (d) detail the assessment of these models and the rank model against existing data regarding the functional form of frequency effects; and (e) report further assessments using contextual diversity, a factor confounded with word frequency. The relationship between the occurrence distribution of words and lexical decision latencies to those words does not appear compatible with the rank hypothesis, undermining the case for serial search models of lexical access. Three transformations of contextual diversity based on extensions of instance models do, however, remain as plausible explanations of the effect.     

Are lexical decisions a good measure of lexical access? The role of word frequency in the neglected decision stage

Three experiments investigated the impact of five lexical variables (instance dominance, category dominance, word frequency, word length in letters, and word length in syllables) on performance in three different tasks involving word recognition: category verification, lexical decision, and pronunciation. Although the same set of words was used in each task, the relationship of the lexical variables to reaction time varied significantly with the task within which the words were embedded. In particular, the effect of word frequency was minimal in the category verification task, whereas it was significantly larger in the pronunciation task and significantly larger yet in the lexical decision task. It is argued that decision processes having little to do with lexical access accentuate the word-frequency effect in the lexical decision task and that results from this task have questionable value in testing the assumption that word frequency orders the lexicon, thereby affecting time to access the mental lexicon. A simple two-stage model is outlined to account for the role of word frequency and other variables in lexical decision. The model is applied to the results of the reported experiments and some of the most important findings in other studies of lexical decision and pronunciation.     

A word’s meaning affects the decision in lexical decision

The influence of an isolated word’s meaning on lexical decision reaction time (RT) was demonstrated through four experiments. Subjects in two experiments made lexical decision judgments, those in a third experiment pronounced the words used in the lexical decision task, and those in a fourth experiment quickly pronounced their first associative response to the words. Differences in lexical access time for the words were measured with the pronunciation task, and differences in meaning were assessed with the association task. Multiple regression analyses of lexical decision RT were conducted using associative RT, pronunciation RT, and other target word properties (printed frequency, length, instance dominance, and number of dictionary meanings) as predictor variables. These analyses revealed a relationship between lexical decision RT and associative RT after the effects of other variables had been partialed out. In addition, word frequency continued to have a significant relationship to lexical decision RT beyond that shared with pronunciation RT and the other variables. The results of these experiments indicate that at least some of the effect of word meaning and word frequency in lexical decision is attributable to a decision stage following lexical access.     

Serial mechanisms in lexical access: the rank hypothesis

There is general agreement that the effect of frequency on lexical access time is roughly logarithmic, although little attention has been given to the reason for this. The authors argue that models of lexical access that incorporate a frequency-ordered serial comparison or verification procedure provide an account of this effect and predict that the underlying function directly relates access time to the rank order of words in a frequency-ordered set. For both group data and individual data, it is shown that rank provides a better fit to the data than does a function based on log frequency. Extensions to a search model are proposed that account for error rates and latencies and the effect of age of acquisition, which is interpreted as an effect of cumulative frequency.     

Joint effects of semantic priming and repetition in a lexical decision task: Implications for a model of lexical access

An experiment is reported which showed that in a lexical decision task semantic priming by a related preceding word and repetition of target words produce additive effects on decision latency. Previous models of lexical access and modifications of them are discussed, and it is argued that some such models predict an interaction of priming and repetition, while others are insufficiently precise to make a prediction. It is suggested that the generality of effects across tasks requiring lexical access must be established and the components of complex effects must be separated before an adequate model can be devised to account for the data.     

Terminating and exhaustive search in lexical access

Two experiments which test predictions derived from the assumption that lexical access involves a search process are reported. In the first experiment, test items must be classified as ambiguous or unambiguous, and in the second experiment, they are classified according to their syntactic properties. In both experiments, it is shown that when the target of the search is a nonexistent entry, an exhaustive search is involved, even though the test items are words. Further, in these conditions, frequency of occurrence is no longer related to decision time, as it is in lexical decision experiments. It is concluded that the search model adequately explains the procedure whereby the most common meaning of a homograph is accessed, but that the less common meaning is accessed in some completely different manner.     

Are there lexicons?

Many models of the processing of printed or spoken words or objects or faces propose that systems of local representations of the forms of such stimuli—lexicons—exist. This is denied by partisans of the distributed-representation connectionist approach to cognitive modelling. An experimental paradigm of key theoretical importance here is lexical decision and its analogue in the domain of objects, object decision. How does each theoretical camp account for our ability to perform these two tasks? The localists say that the tasks are done by matching or failing to match a stimulus to a local representation in a lexicon. Advocates of distributed representations often do not seek to explain these two tasks; however, when they do, they propose that patterns of activation evoked in a semantic system can be used to discriminate between words and nonwords, or between real objects and false objects. Therefore the distributed-representation account of lexical and object decision tasks predicts that performance on these tasks can never be normal in patients with an impaired semantic system, nor in patients who cannot access semantics normally from the stimulus domain being tested. However, numerous such patients have been reported in the literature, indicating that semantic access is not needed for normal performance on these tasks. Such results support the localist form of modelling rather than the distributed-representation approach.     

Attentional resource demands of visual word recognition in naming and lexical decisions.

Attentional demands of lexical access were assessed with dual-task methodology. Subjects performed an auditory probe task alone (single-task) or combined (dual-task) with either a lexical decision or a naming task. In Experiment 1, probe performance showed a decrement from single- to dual-task conditions during recognition of words in both lexical decision and naming tasks. In addition, decrements in probe performance were larger during processing of low-frequency compared with high-frequency words in both of the word recognition tasks. Experiment 2 showed that the time course of frequency-sensitive demands was similar across lexical decision and naming tasks and that attention is required early in the word recognition sequence. The results support the assumption that lexical access is both frequency sensitive and attention demanding.     

Additive and interactive effects on response time distributions in visual word recognition

Across 3 different word recognition tasks, distributional analyses were used to examine the joint effects of stimulus quality and word frequency on underlying response time distributions. Consistent with the extant literature, stimulus quality and word frequency produced additive effects in lexical decision, not only in the means but also in the shape of the response time distributions, supporting an early normalization process that is separate from processes influenced by word frequency. In contrast, speeded pronunciation and semantic classification produced interactive influences of word frequency and stimulus quality, which is a fundamental prediction from interactive activation models of lexical processing. These findings suggest that stimulus normalization is specific to lexical decision and is driven by the task's emphasis on familiarity-based information.     

Single- versus dual-process models of lexical decision performance: insights from response time distributional analysis

This article evaluates 2 competing models that address the decision-making processes mediating word recognition and lexical decision performance: a hybrid 2-stage model of lexical decision performance and a random-walk model. In 2 experiments, nonword type and word frequency were manipulated across 2 contrasts (pseudohomophone-legal nonword and legal-illegal nonword). When nonwords became more wordlike (i.e., BRNTA vs. BRANT vs. BRANE), response latencies to nonwords were slowed and the word frequency effect increased. More important, distributional analyses revealed that the Nonword Type = Word Frequency interaction was modulated by different components of the response time distribution, depending on the specific nonword contrast. A single-process random-walk model was able to account for this particular set of findings more successfully than the hybrid 2-stage model.     

Length effect in reading and lexical decision: evidence from skilled readers and a developmental dyslexic participant

A number of experimental data have shown that naming latency increases with length for pseudo-words but not for frequent real words. Different interpretations have been proposed by current models of reading to account for such a length effect. The aim of the present study was to assess the impact of lexicality on length effect in both the reading and lexical decision tasks. For this purpose, skilled readers were asked to either name or make a lexical decision on words and pseudo-words differing in length from one to three syllables. Skilled readers' results show that length effect is modulated by lexicality in the reading task but no length effect was found in the lexical decision task. The tasks were further proposed to a well-compensated dyslexic participant who exhibited a visual attentional disorder in the absence of any associated phonological problems. A length effect on RTs was found for both words and pseudo-words in lexical decision but naming latencies were affected by length for the pseudo-words only. The present results largely conform to the predictions of the ACV98 model of reading. They are not compatible with the PDP models of reading and can only be partially accounted for by dual route models.     

Reaction time methodology and the aphasic patient: a reply to Hagoort (1988)

P. Hagoort (1988, Brain and Language, 36, 335-348) suggests that the results of an experiment conducted with aphasic patients reported by W. Milberg, S. E. Blumstein, and B. Dworetzky (1987, Brain and Language, 31, 138-150) are "difficult to interpret in terms of a selective access model for normal processing because of a confound caused by the use of repeated lexical decision targets." We argue that the use of repeated lexical decision targets and other methodological compromises are necessary for using the lexical decision reaction time paradigm with aphasic patients. Furthermore, the minor differences between our data and those presented by Hagoort do not undermine our claims concerning semantic processing in different subgroups of aphasics, nor our comparisons of these patients to normal subjects.     

Age of acquisition in face categorisation: is there an instance-based account?

An instance-based model of the effects of age of acquisition (AoA) is offered and derived predictions are considered. A face-categorisation experiment is reported which tested these predictions. Nine participants made speeded-categorisation judgements of 185 faces from two TV shows. Ratings were given for the faces' frequency of occurrence in the shows and, for each face, the length of time the character was in the show and the time since they had left was found. Regression analyses were used to find how the factors affected the speed of categorisation. Speed of categorisation was found to follow a negative power function of the frequency and time on the show and a positive power function of the time since they had left. This was supportive of the instance-based account and suggests that effects of AoA with words and objects can be explained in terms of cumulative frequency.     

Distinguishing common and task-specific processes in word identification: a matter of some moment?

The same 500 words were presented in 6 different word identification tasks (Experiment 1: lexical decision, semantic categorization, and 3 speeded naming tasks; Experiment 2: delayed naming). Reaction time (RT) distributions were estimated for each task and analyses tested for the effects of word frequency and animacy on various parameters of the RT distribution. Low frequency words yielded more skewed distributions than high frequency words in all tasks except delayed naming. The differential skew was most marked for tasks that required lexical discrimination. The semantic categorization task yielded highly skewed distributions for all words, but the word frequency effect was due to shifts in the location of the RT distributions rather than changes in skew. The results are used to evaluate the relative contributions of a common lexical access process and task-specific processes to performance in lexical discrimination and naming tasks.     

Lexical access and the spelling-to-sound regularity effect

A number of investigators have reported that words that follow spelling-to-sound rules can be recognized faster than words that violate such rules (the “regularity” effect). On occasion, the absence of a regularity effect is reported, however. The first two experiments of the present paper report that a regularity effect can be obtained in a lexical decision task with word sets that previously have been reported not to produce such an effect, when consideration is given to the consistency or inconsistency of the pronunciations of each word’s visually similar ‘neighbors” Subsequent experiments demonstrated that the obtained regularity effect does not vary as a function of mixed- vs. single-case presentation (Experiment 3) or visual quality (Experiment 4) in a lexical decision task. These results are explained in terms of Glushko’s (1979) activation and synthesis model of lexical access. It is argued that the obtained results are incompatible with traditional dual process models of lexical access (which incorporate separate visual and phonological pathways and spelling-to-sound rules) and fully compatible with Glushko’s model. It is concluded that spelling-to-sound regularity is not a property of a word in isolation, but rather a property of a word in the context of visually similar words that are activated in the course of recognition.     

Morphological decomposition and the reverse base frequency effect

If recognition of a polymorphemic word always takes place via its decomposition into stem and affix, then the higher the frequency of its stem (i.e., base frequency) the easier the lexical decision response should be when frequency of the word itself (i.e., surface frequency) is controlled. Past experiments have demonstrated such a base frequency effect, but not under all circumstances. Thus, a dual pathway notion has become dominant as an account of morphological processing whereby both decomposition and whole-word access is possible. Two experiments are reported here that demonstrate how an obligatory decomposition account can handle the absence of base frequency effects. In particular, it is shown that the later stage of recombining the stem and affix is harder for high base frequency words than for lower base frequency words when matched on surface frequency, and that this can counterbalance the advantage of easier access to the higher frequency stem. When the combination stage is crucial for discriminating the word items from the nonword items, a reverse base frequency effect emerges, revealing the disadvantage at this stage for high base frequency words. Such an effect is hard for the dual-pathway account to explain, but follows naturally from the idea of obligatory decomposition.     

Lexical status of inflectional and derivational suffixes: Evidence from Finnish

The present study tested whether Finnish inflectional or derivational suffixes have lexical access units separate from word roots. In three lexical decision experiments, pseudowords carrying a case-inflection required significantly longer rejection times than nonaffixed pseudowords. This suggests that case-inflections have separate lexical access units. Similar effects were obtained for productive derivational suffixes, too. A specific lexical architecture, being able to account for both the present pseudoword results as well as earlier ones obtained with real Finnish nouns, is proposed.     

Exploring the additive effects of stimulus quality and word frequency: The influence of local and list-wide prime relatedness

Stimulus quality and word frequency produce additive effects in lexical decision performance, whereas the semantic priming effect interacts with both stimulus quality and word frequency effects. This pattern places important constraints on models of visual word recognition. In Experiment 1, all three variables were investigated within a single speeded pronunciation study. The results indicated that the joint effects of stimulus quality and word frequency were dependent upon prime relatedness. In particular, an additive effect of stimulus quality and word frequency was found after related primes, and an interactive effect was found after unrelated primes. It was hypothesized that this pattern reflects an adaptive reliance on related prime information within the experimental context. In Experiment 2, related primes were eliminated from the list, and the interactive effects of stimulus quality and word frequency found following unrelated primes in Experiment 1 reverted to additive effects for the same unrelated prime conditions. The results are supportive of a flexible lexical processor that adapts to both local prime information and global list-wide context.     

Effect of backward priming on word recognition in single-word and sentence contexts

Backward priming was investigated under conditions similar to those used in lexical ambiguity research. Subjects received prime-target word pairs that were associated either unidirectionally (BABY-STORK) or bidirectionally (BABY-CRY). In the first experiment, targets were presented 500 ms following the onset of visual primes, and subjects made naming or lexical decision responses to the targets. Forward priming was obtained in all conditions, while backward priming (i.e., priming for pairs in which there was a unidirectional target-to-prime association, as in BABY-STORK) occurred only with lexical decision. In the second experiment, primes were presented auditorily, either in isolation or in a sentence. Targets followed the offset of the primes either immediately or after 200 ms. Backward priming occurred with both response tasks, but only when the prime was an isolated word. In addition, backward priming decreased over time with the naming task, but not with lexical decision. These results suggest that the locus of the backward priming effect is different for the two response tasks. Further, the lack of a backward priming effect with sentence contexts suggests that backward priming cannot account for the demonstrations of multiple access in the lexical ambiguity literature. These results, therefore, support a context-independent view of lexical access.