Interactive Activation Model Research Articles

The Effect of the Balance of Orthographic Neighborhood Distribution in Visual Word Recognition

The present study investigated whether the balance of neighborhood distribution (i.e., the way orthographic neighbors are spread across letter positions) influences visual word recognition. Three word conditions were compared. Word neighbors were either concentrated on one letter position (e.g.,nasse/basse-lasse-tasse-masse) or were unequally spread across two letter positions (e.g.,pelle/celle-selle-telle-perle), or were equally spread across two letter positions (e.g.,litre/titre-vitre-libre-livre). Predictions based on the interactive activation model [McClelland & Rumelhart (1981). Psychological Review, 88, 375-401] were generated by running simulations and were confirmed in the lexical decision task. Data showed that words were more rapidly identified when they had spread neighbors rather than concentrated neighbors. Furthermore, within the set of spread neighbors, words were more rapidly recognized when they had equally rather than unequally spread neighbors. The findings are explained in terms of activation and inhibition processes in the interactive activation framework.

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.

Analogical effects on linking elements in German compound words

This paper examines whether the selection of linking elements for novel German compounds can be better explained in terms of a single or a dual-route model. Previous studies had focused on the predictability of linking elements by rules. We investigate a single-route model by focusing on the paradigmatic analogical effect of the compounds sharing the left (right) constituent with the target compound, i.e., the left (right) constituent family. A production experiment reveals an effect of the left, but not of the right constituent family. Simulation studies of the responses, using a computational model of paradigmatic analogy, show that the left constituent and its phonological and morphological properties (rime, gender, and inflectional class) simultaneously codetermine the selection of linking elements. We show how these results can be accounted for by a single-route approach, and we outline a symbolic interactive activation model that merges the factors into one psycholinguistically motivated processing mechanism.

Effectiveness of contextual repetition priming treatments for anomia depends on intact access to semantics

Naming of two semantically impaired aphasic patients was treated with the contextual repetition priming technique, which involves repeated repetition of names of pictures that are related semantically, phonologically, or are unrelated. Our previous studies using this technique have suggested that patients with impaired access to lexical-semantic representations benefit in the short-term from this treatment technique, but show no long-term improvement in naming. In contrast, patients with good access to semantics show short- and long-term benefits from this treatment. Here we report two studies of treatment for two individuals with aphasia affecting access to lexical semantics and anomia but spared access to input and output phonology and spared conceptual semantics. We predicted that they would show short-term facilitation from the contextual priming, but no long-term improvements in naming. The results confirmed the prediction. An account of this pattern is offered within the framework of an interactive activation model of word retrieval. Additionally, we discuss alternative techniques for addressing naming deficits when access to semantics is impaired.

FIRST LANGUAGE ACTIVATION DURING SECOND LANGUAGE LEXICAL PROCESSING: An Investigation of Lexical Form, Meaning, and Grammatical Class

This study places the predictions of the bilingual interactive activation model (Dijkstra & Van Heuven, 1998) and the revised hierarchical model (Kroll & Stewart, 1994) in the same context to investigate lexical processing in a second language (L2). The performances of two groups of native English speakers, one less proficient and the other more proficient in Spanish, were compared on translation recognition. In this task, participants decided whether two words, one in each language, are translation equivalents. The items in the critical conditions were not translation equivalents and therefore required a “no” response, but were similar to the correct translation in either form or meaning. For example, for translation equivalents such as cara-face, critical distracters included (a) a form-related neighbor to the first word of the pair (e.g., cara-card), (b) a form-related neighbor to the second word of the pair, the translation equivalent (cara-fact), or (c) a meaning-related word (cara-head). The results showed that all learners, regardless of proficiency, experienced interference for lexical neighbors and for meaning-related pairs. However, only the less proficient learners also showed effects of form relatedness via the translation equivalent. Moreover, all participants were sensitive to cues to grammatical class, such that lexical interference was reduced or eliminated when the two words of each pair were drawn from different grammatical classes. We consider the implications of these results for L2 lexical processing and for models of the bilingual lexicon.The writing of this article was supported in part by NSF Doctoral Enhancement Grant BCS-0111733 to Gretchen Sunderman and Judith F. Kroll, and by NSF grants BCS-0111734 and BCS-0418071 and NIH grant RO1MH62479 to Judith F. Kroll. We thank Maya Misra for advice on computing measures of orthographic similarity and Rachel Varra and Asha Persaud for research assistance. We also thank the anonymous SSLA reviewers for their helpful comments.

The nature of the syllabic neighbourhood effect in French

We investigated whether and how sublexical units such as phonological syllables mediate access to the lexicon in French visual word recognition. To do so, two lexical decision task (LDT) experiments examined the nature of the syllabic neighbourhood effect. In Experiments 1a and b, the number of higher frequency syllabic neighbours was manipulated while controlling for the first bigram. The results failed to show a pure syllabic neighbourhood effect. In Experiments 2a and b, syllabic neighbourhood and bigram frequency were factorially manipulated. The interaction showed that the syllabic neighbourhood effect was inhibitory when bigram frequency was high, whereas it was facilitatory when bigram frequency was low. Similar patterns of results were found in both the yes/no (Experiments 1a and 2a) and go/no-go LDTs (Experiments 1b and 2b), so varying task requirements of the lexical decision did not influence the effect. These findings are discussed in the context of parallel distributed processing and interactive-activation models, and suggest that orthographic redundancy properties contribute to the influence of phonological syllables.

Masked inhibitory priming in English: Evidence for lexical inhibition.

Predictions derived from the interactive activation (IA) model were tested in 3 experiments using the masked priming technique in the lexical decision task. Experiment 1 showed a strong effect of prime lexicality: Classifications of target words were facilitated by orthographically related nonword primes (relative to unrelated nonword primes) but were inhibited by orthographically related word primes (relative to unrelated word primes). Experiment 2 confirmed IA's prediction that inhibitory priming effects are greater when the prime and target share a neighbor. Experiment 3 showed a minimal effect of target word neighborhood size (N) on inhibitory priming but a trend toward greater inhibition when nonword foils were high-N than when they were low-N. Simulations of 3 different versions of the IA model showed that the best fit to the data is produced when lexical inhibition is selective and when masking leads to reset of letter activities.

Hierarchical schemas and goals in the control of sequential behavior.

Traditional accounts of sequential behavior assume that schemas and goals play a causal role in the control of behavior. In contrast, M. Botvinick and D. C. Plaut argued that, at least in routine behavior, schemas and goals are epiphenomenal. The authors evaluate the Botvinick and Plaut account by contrasting the simple recurrent network model of Botvinick and Plaut with their own more traditional hierarchically structured interactive activation model (R. P. Cooper & T. Shallice, 2000). The authors present a range of arguments and additional simulations that demonstrate theoretical and empirical difficulties for both Botvinick and Plaut's model and their theoretical position. The authors conclude that explicit hierarchically organized and causally efficacious schema and goal representations are required to provide an adequate account of the flexibility of sequential behavior.

Processes of Similarity Judgment

Similarity underlies fundamental cognitive capabilities such as memory, categorization, decision making, problem solving, and reasoning. Although recent approaches to similarity appreciate the structure of mental representations, they differ in the processes posited to operate over these representations. We present an experiment that differentiates among extant structural accounts of similarity in their ability to account for patterns of similarity ratings. These data pose a challenge for transformation-based models and all but one mapping-based model, the Similarity as Interactive Activation and Mapping (SIAM) model of similarity.

Chasing concepts during design: A photo shoot from the field of architecture

This paper examines the role of design concepts in a modus operandi as opposed to a modus operatum, which is how their generation, as it unfolds over time, is perceived by someone involved in designing instead of with the hindsight of being finished. To this end, the interactive activation and competition model, commonly used to model the retrieval of information from stored knowledge of individual exemplars, is applied in the context of architectural design. Parsing the data of an architect's think-aloud protocol through this model at successive points in the design process results in a photo shoot of a design concept “under construction.” This allows for the start of appreciating and accounting for the highly elusive character of concepts during design.

Some empirical tests of an interactive activation model of eye movement control in reading

This paper describes some empirical tests of an interactive activation model of eye movement control in reading (the “Glenmore” model). Qualitatively, the Glenmore model can account within one mechanism for preview and spillover effects, regressions, progressions, and refixations. It decouples the decision about when to move the eyes from the word recognition process. The time course of activity in a fixate centre (FC) determines the triggering of a saccade. The other main feature of the model is the use of a saliency map that acts as an arena for the interplay of bottom-up visual features of the text, and top-down lexical features. These factors combine to create a pattern of activation that selects one word as the saccade target. Even within the relatively simple framework proposed here, a coherent account can be provided for a range of eye movement control phenomena that have hitherto proved problematic to reconcile. The paper examines the performance of the model compared to data gathered in an empirical study of subjects reading a German text. The quantitative fit of the model, while reasonable, highlighted some limitations in the model that will need to be addressed in future versions.

Effets de distribution du voisinage orthographique et d'amorçage par répétition masqué.

This research centres on the effect that the orthographic neighbourhood has in the visual recognition of words. Specifically, we studied to what extent orthographic neighbourhood distribution, that is, the number of letter positions allowing formation of at least one neighbour (Pugh, Rexer, Peter, & Katz, 1994), influences the masked repetition priming effect. In a previous study (Mathey, Robert, & Zagar, 2004), interaction between neighbourhood distribution and orthographic priming was obtained in the lexical decision task. The Interactive Activation Model (IA; McClelland & Rumelhart, 1981) simulated this interaction. With the orthographic priming effect modified for distribution of the neighbourhood of target words, it was necessary to study whether the repetition priming effect also varied as a function of this indicator. Studying this interaction presents a major theoretical issue in specifying the activating and inhibiting processes presented in the IA model. Simulations were produced to obtain precise model predictions regarding the neighbourhood distribution effect in a repetitive priming situation for our experimental material. Target words all had two neighbours that were most frequent. These neighbours were isolated, that is, distributed over two letter positions (e.g.: TAUX/faux-toux), or associated, i.e., concentrated on one single position (e.g., SEAU/beau-peau). Targets were preceded by an identical priming (repetitive priming; e.g.: seau-SEAU) or by controlled priming (e.g., &&&&-SEAU). The simulation results obtained using the IA model show the facilitating effects of neighbourhood distribution and repetitive priming, but no interaction between these factors. The experimental results obtained in a lexical decision task confirm these predictions. Thus, the empirical data replicate the neighbourhood distribution's facilitating effect (Mathey & Zagar, 2000) as well as the facilitating effect of masked repetition (Forster & Davis, 1984). Finally, the most interesting result is that the facilitating effect of repetition is comparable for target words with associated neighbours and target words with isolated neighbours. An explanation of the combined effects of the orthographic neighbourhood and orthographic masked repetition priming, integrating data from literature as well as from the current study, is proposed within the framework of the IA model.

Zooming into L2: Global language context and adjustment affect processing of interlingual homographs in sentences

In a semantic priming study, we investigated the processing of German–English homographs such as gift (German = “poison”, English = “present”) in sentence contexts using a joint reaction time (RT)/event-related brain potential (ERP) measure. Native German speakers with intermediate or advanced knowledge of English ( N = 48) performed an all-L2 (English) experiment where sentences such as “The woman gave her friend an expensive gift” (control prime: item) were presented, followed by targets (i.e., boss) for lexical decision. To test the role of global task effects during sentence processing, we presented half the participants ( N = 24) with a 20-min silent film narrated in German and half ( N = 24) with the film in English before the experiment. To address the development of task effects over time, we analyzed the first and second blocks of the experiment. The results showed a significant interaction between semantic priming, movie version, and block in both the RTs and ERPs: there was significant semantic priming in the RTs and modulations in the N200 and N400 components only for participants who viewed the German movie, and only during the first block. Results suggest that in an all-L2 sentence task with L2 pre-task priming (English film), decision thresholds are raised high enough to eliminate measurable influence of the L1 on the L2. Despite identical material, participants who viewed the German film had to adjust, or zoom in, to the all-L2 task. Implications of this zooming in process in are discussed in terms of the recent Bilingual Interactive Activation (BIA+) model of bilingual word recognition [T. Dijkstra, W.J.B. Van Heuven, The architecture of the bilingual word recognition system: from identification to decision, Bilingualism: Lang. Cogn. 5 (2002) 175–197].

Letter detection for homographs with different meanings in different language texts

Tests of inter-lingual homographs that have different meanings across two languages support models postulating initial non-selective access to competing language representations, e.g. Bilingual Interactive Activation (BIA) model. Most such research assessed inter-lingual homographs in the absence of connected text. Here a letter detection paradigm was used that required subjects to detect letters in words in connected text. Prior work with this paradigm suggested that readers respond to only one interpretation of an intra-lingual homograph when detecting letters. Three experiments described here indicate that letter detection patterns to inter-lingual homographs are similar, i.e. detection reflects only a context appropriate interpretation. However, the demonstration that text role, text cohesiveness and bilingual fluency affect inter-lingual letter detection (Experiments 1 and 2), and that word role affects detection even though target frequency is constant across inter-lingual meanings (Experiment 3) indicates that selectivity is in response to post-lexical processes. Thus, results are seen as compatible with tenets of the BIA model.

Perseverations and Anticipations in Aphasia: Primed Intrusions From the Past and Future

In this article, we review an account of movement errors that is based on a model of serial order and an interactive spreading activation model of word production. This account makes two claims. First, anticipations and perseverations arise from malfunctions of a mechanism that maintains serial order in speech production and that has three components: (1) a means to turn off past utterances, (2) a means to activate the present utterance (3) a means to prime the future utterance. Second, the probabilities of such movement errors are further determined by the same variables that influence the production of ordinary word and sound substitutions, such a frequency and feature overlap. We review relevant studies of movement errors produced by aphasic and nonaphasic individuals and make suggestions for further investigation of these phenomena.

Immediate and short‐term effects of contextual priming on word retrieval in aphasia

Background: Many therapy techniques for word retrieval disorders use some form of priming to improve access to words. Priming can facilitate or interfere with naming under different circumstances. We examined effects of priming when combined with semantic or phonological context (training words in groups that are semantically or phonologically related) and how these effects interact with the type of naming impairment (semantically or phonologically based).Aims: We addressed three questions (1) Are word retrieval impairments differentially sensitive to priming with semantic or phonological contexts? (2) Would such differences be systematically related to deficits of semantic versus phonological processing? (3) Do effects of priming evolve from immediate interference to short‐term facilitation, as predicted by an interactive activation model of word retrieval?Methods & Procedures: A total of 11 chronic English‐speaking aphasic subjects with varied types of aphasia participated in this experiment. Background measures of semantic and phonological processing ability were administered to determine the nature of each subject's naming impairment. The experiment involved one‐session facilitation treatments for each of three context conditions (semantic, phonological, and unrelated), plus three replications (nine subjects) or one replication (two subjects). Ten pictures in each condition were tested before and after treatment. Five pictures were trained and five served as controls. Participants repeated the name of each picture four times (repetition priming) and then attempted to name each picture individually (naming probe). Repetition priming and naming probes were repeated eight times. We used McNemar tests to compare rates of correct responses before and after priming, and chi square analyses of correct responses and contextual errors on naming probes obtained during the priming sessions.Outcome & results: Our predictions were borne out in the data. Participants varied in their sensitivity to the semantic and phonological contexts. The error data suggest that interference during training is more likely when the context (semantic or phonological) and underlying source of the word processing impairment (semantic or phonological) match. Additionally, we found two sequential effects of contextual priming: immediate interference followed short‐term facilitation.Conclusions: These data have theoretical implications regarding the time course of priming effects, but also have important clinical implications. The present contextual priming procedure is relatively short and could be used as a predictor of performance patterns in a long‐term treatment protocol that uses this approach or other tasks that employ priming.

Can CANISO activate CASINO? Transposed-letter similarity effects with nonadjacent letter positions

Nonwords created by transposing two adjacent letters (i.e., transposed-letter (TL) nonwords like jugde) are very effective at activating the lexical representation of their base words. This fact poses problems for most computational models of word recognition (e.g., the interactive-activation model and its extensions), which assume that exact letter positions are rapidly coded during the word recognition process. To examine the scope of TL similarity effects further, we asked whether TL similarity effects occur for nonwords created by exchanging two nonadjacent letters (e.g., caniso- CASINO) in three masked form priming experiments using the lexical decision task. The two nonadjacent transposed letters were consonants in Experiment 1 (e.g., caniso- CASINO), vowels in Experiment 2 ( anamil-ANIMAL) and both consonants and vowels in Experiment 3. Results showed that nonadjacent TL primes produce priming effects (in comparison to orthographic controls, e.g., caviro- CASINO), however, only when the transposed letters are consonants. In a final experiment we examined latencies for nonwords created by nonadjacent transpositions of consonants versus vowels in a lexical decision task. Both types of nonwords produced longer latencies than matched controls, with consonant TL nonwords being more difficult than vowel TL nonwords. The implications of these findings for models having “position-specific” coding schemes as well as for models proposing alternative coding schemes are discussed.

Recognizing cognates and interlingual homographs: effects of code similarity in language-specific and generalized lexical decision.

In four experiments, we investigated how cross-linguistic overlap in semantics, orthography, and phonology affects bilingual word recognition in different variants of the lexical decision task. Dutch-English bilinguals performed a language-specific or a generalized lexical decision task including words that are spelled and/or pronounced the same in English and in Dutch and that matched one-language control words from both languages. In Experiments 1 and 3, "false friends" with different meanings in the two languages (e.g., spot) were presented, whereas in Experiments 2 and 4 cognates with the same meanings across languages (e.g., film) were presented. The language-specific Experiments 1 and 2 replicated and qualified an earlier study (Dijkstra, Grainger, & Van Heuven, 1999). In the generalized Experiment 3, participants reacted equally quickly on Dutch-English homographs and Dutch control words, indicating that their response was based primarily on the fastest available orthographic code (i.e., Dutch). In Experiment 4, cognates were recognized faster than English and Dutch controls, suggesting coactivation of the cognates' semantics. The nonword results indicate that the bilingual rejection procedure can, to some extent, be language specific. All results are discussed within the BIA+ (bilingual interactive activation) model for bilingual word recognition.

Testing predictions of the interactive activation model in recovery from aphasia after treatment

This paper presents preliminary results of pre- and post-treatment error analysis from an aphasic patient with anomia. The Interactive Activation (IA) model of word production ( Dell, Schwartz, Martin, Saffran, & Gagnon, 1997) is utilized to make predictions about the anticipated changes on a picture naming task and to explain emerging patterns. Error patterns are viewed in light of two putative mechanisms within the IA network, connection strength, and rate of decay. The results suggest that while these mechanisms can successfully account for the magnitude and order of specific errors (e.g., predominance of semantic paraphasias and nonwords followed by formal paraphasias under the conditions of weak connections), the treatment may have influenced the occurrence/absence of some error types.

Syllable-frequency effects in visual word recognition: evidence from ERPs.

A lexical decision experiment was conducted while event related potentials (ERPs) were recorded. The word frequency and the first syllable frequency of each word were manipulated. Results showed that, while high frequency words produced less negative amplitudes in the N400 time window than low frequency words, the inverse pattern was found for syllable frequency. Words containing high frequency syllables produced more negative amplitudes than words containing low frequency syllables. Importantly, a significant syllable frequency effect was also obtained at the P200 time window. The results are interpreted in the framework of an interactive activation model, in which high frequency syllables produce the initial activation of a larger number of lexical candidates during the analysis of orthographic or phonological representations, which have to be inhibited later to allow the identification of a unique word. These findings support the idea that, at least in languages with clear syllabic boundaries, syllables are functional sublexical units during visual word recognition.