Introduction

WEAVER++ (Word Encoding by Activation and VERification) is a computational model designed to explain how we, homo loquens, plan the production of spoken words (also called word finding, word retrieval, or lexical access); how the planning relates to other cognitive abilities, including perception, action, memory, thinking, and in particular, attention; and how word planning is disrupted by brain damage, due to stroke or neurodegenerative disease. The model tries to make explicit the many strands of declarative knowledge ("knowing that") that must be woven together using procedural knowledge ("knowing how") and the different mixtures of strands used in different word production tasks. The model falls into the general class of hybrid models of human performance in that it combines a declarative associative network and procedural rule system with retrieval from declarative memory through spreading activation and activation-based rule triggering (cf. ACT-R of John Anderson and colleagues). The distinction between declarative and procedural knowledge in word planning is reminiscent of the distinction between mental contents and acts advanced by Oswald Külpe based on the work of his Würzburg group in the early 1900s, and it is supported by accumulating empirical evidence on language production (e.g., Pim Levelt's 1989 Speaking; see also the work of Michael Ullman and colleagues). Milestones in the history of the distinction include Brenda Milner’s (1962) study on mirror-drawing by patient H.M., computational work in Artificial Intelligence in the 1970s (e.g.,Terry Winograd, 1975), and Neil Cohen and Larry Squire’s (1980) study on mirror-reading by amnesic patients. The WEAVER++ model plans spoken words by activating, selecting, and connecting (weaving together) types of verbal information.

WEAVER++ gives detailed accounts of response time (RT) findings on spoken word production, obtained within a research tradition originating with Frans Donders, James McKeen Cattell, and John Ridley Stroop. Click on the icons above for more information on mental chronometry and for an interactive demo of the color-word Stroop test. WEAVER++ has also been applied to eye tracking and neuroimaging findings (i.e., electrophysiological as well as hemodynamic evidence). Moreover, the model has been applied to evidence on response accuracy in impairments, obtained within a research tradition originating with aphasiologists like Carl Wernicke and Arnold Pick. The neurocognitive extension WEAVER++/ARC synthesizes behavioral psycholinguistic, functional neuroimaging, tractography, and aphasiological evidence. The goal is to provide unified computational explanations for a wide range of relevant empirical findings on spoken word planning (word finding, word retrieval, lexical access) in health and disease. See for example:

Roelofs, A. (2014). A dorsal-pathway account of aphasic language production: The WEAVER++/ARC model. Cortex, 59, 33-48. Article (PDF 1339K)

Roelofs, A. (2018). A unified computational account of cumulative semantic, semantic blocking, and semantic distractor effects in picture naming. Cognition, 172, 59-72. Article (PDF 574K)

Roelofs, A. (2022). A neurocognitive computational account of word production, comprehension, and repetition in primary progressive aphasia. Brain and Language, 227, 105094. Article (PDF 2399K)

The model has been developed for Germanic languages like Dutch and English, but also model versions for Mandarin Chinese and Japanese have been made.

 

 

 

Roelofs, A. (2015). Modeling of phonological encoding in spoken word production: From Germanic languages to Mandarin Chinese and Japanese. Japanese Psychological Research, 57, 22-37. Article (PDF 513K)

 

 

The WEAVER++ model takes position regarding a number of big questions in the history of psychology, such as the issue of imageless thought (i.e., the question of whether abstract concept representations exist or only perceptual and motor representations, nowadays called embodied cognition) and the issue of mental set (i.e., how goals direct processing, which is through procedural if-then rules, according to the model). For a description of the roots of these issues in psychological science, see the classic book Experimental Psychology by Robert Woodworth (1938).

 

 

Horizontal and vertical threads

 

Accumulating evidence suggests that linguistic processes underlying the planning of words cannot happen without paying some form of attention. Or as Adolf Kussmaul put it in Die Störungen der Sprache (1877), language processes proceed semi-automatically ("halb automatisch").


In their classic article "Attention to action: Willed and automatic control of behavior", Don Norman and Tim Shallice (1986) made a distinction between "horizontal threads" and "vertical threads" in the control of human performance. Horizontal threads are strands of processing that map perceptions onto actions and vertical threads are attentional influences on these mappings. Characteristics of performance arise from interactions between horizontal and vertical threads. WEAVER++ implements specific claims about how the horizontal and vertical threads are woven together in language performance.

A central claim embodied by WEAVER++ is that the attentional control of language performance is achieved through condition-action rules (cf. EPIC of David Meyer and David Kieras) rather than purely associatively. WEAVER++'s lexical network is accessed by spreading activation while the condition-action rules determine what is done with the activated lexical information depending on the task. When a goal is specified in working memory, the processing of the system is focussed on those rules that include the goal among their conditions.

WEAVER++ plans words by incrementally extending verbal goals, i.e., lemmas are selected for selected lexical concepts, morphemes for selected lemmas, segments for selected morphemes, and syllable motor programs for selected and syllabified segments, whereby the syllabification of segments proceeds incrementally from the beginning of a word to its end. The idea of incrementality in language production originates with Wilhelm Wundt. Furthermore, WEAVER++ is an attempt at incremental or cumulative modeling, i.e., an attempt to extend a model in new directions and to new phenomena by building on earlier modeling results.

 

Multiple memory systems

 

The model makes explicit how three major memory systems of the brain, i.e., declarative, procedural, and working memory, interact during spoken word planning. When a goal (e.g., to name a picture) is temporarily maintained in working memory, information to achieve the goal is retrieved from long-term declarative memory through the application of condition-action rules from long-term procedural memory. Below, this is further explained. For comprehensive discussions of the brain systems for declarative, procedural, and working memory, see the books From Conditioning to Conscious Recollection: Memory Systems of the Brain by Howard Eichenbaum and Neil Cohen (2001) and The Cognitive Neuroscience of Memory: An Introduction by Howard Eichenbaum (2012).

 

 

Varieties of attention

 

Attention is an umbrella term covering a number of abilities, such as alerting (brief or sustained), orienting (with or without gaze shifting), and executive control, also called executive attention, attentional control, or cognitive control (see the work of Michael Posner and colleagues, reviewed by Posner in his 2012 book Attention in a Social World). Moreover, executive control consists of a number of abilities, including updating, inhibiting, and shifting (see the work of Akira Miyake and colleagues). These executive abilties are related to the more traditional notions of divided attention (between tasks), selective attention, and attention switching (between tasks). Selective attention includes the ability to resolve conflict among competing types of information, which can be achieved by enhancing the activation of targets, inhibiting competitors, or both.

 

Moreover, in performing tasks requiring selective attention, such as the color-word Stroop task (e.g., name the ink color of the word GREEN; say "red") or the picture-word interference task (e.g., name a picture of a cat with the word DOG superimposed; say "cat"), WEAVER++ may employ two additional mechanisms of selective attention, referred to as "stimulus set" and "response set" by Donald Broadbent (Decision and Stress book, 1971). Stimulus set ("filtering") refers to selection on the basis of a perceptual attribute, such as spatial location, color, shape, or temporal order. Response set refers to selection on the basis of the vocabulary of allowable responses. Task performance may require one or both of these mechanisms of selective attention:

Roelofs, A. (2003). Goal-referenced selection of verbal action: Modeling attentional control in the Stroop task. Psychological Review, 110, 88-125. Article (PDF 585K)

 

From Wernicke to WEAVER++

In the early days of experimental psychology, Wilhelm Wundt (1880, 1902/1904, Principles of Physiological Psychology book) criticized the now classic, associative Wernicke-Lichtheim model of word production and perception by arguing that the retrieval of words from memory is an active goal-driven process rather than a passive associative process, as held by the model. According to Wundt, an attentional process in the frontal lobes of the human brain controls a word perception and production network located in perisylvian brain areas, described by the Wernicke-Lichtheim model. WEAVER++ builds in many respects on the Wernicke-Lichtheim model, but also addresses Wundt's critique by implementing assumptions on how the production-perception network is controlled. Characteristics of language performance, such as latencies and errors, arise from interactions between the lexical network and the attentional control system. For example, patterns of speech errors by aphasic and nonaphasic speakers seem determined, at least in part, by self-monitoring, which is an important attentional control function:

Roelofs, A. (2011). Modeling the attentional control of vocal utterances: From Wernicke to WEAVER++. In J. Guendouzi, F. Loncke, & M. J. Williams (Eds.), The Handbook of Psycholinguistic and Cognitive Processes: Perspectives in Communication Disorders (pp. 189-207). Hove, UK: Psychology Press. Article (PDF 1464K)

Roelofs, A. (2004). Error biases in spoken word planning and monitoring by aphasic and nonaphasic speakers: Comment on Rapp and Goldrick (2000). Psychological Review, 111, 561-572. Article (PDF 150K)

 

 

WEAVER++/ARC on poststroke aphasia

An extension of WEAVER++ called WEAVER++/ARC (for WEAVER++ Arcuate Repetition and Conversation) synthesizes behavioral psycholinguistic, functional neuroimaging, tractographic, and aphasiological evidence. The model provides a computational implementation of a dorsal pathway view on aphasic language production. According to this view, conceptually driven word production, as involved in spontaneous speech and naming, as well as speech repetition are underpinned by a dorsal white-matter fiber tract in the human brain, called the arcuate fasciculus (AF).

The AF arches around the Sylvian fissure, running from temporal to frontal cortex. The AF is much larger in humans than in nonhuman primates, and the cortical terminations of the AF are strongly modified, suggesting a human specialization that is relevant to the evolution of language (see the work of James Rilling and colleagues). It has long been assumed that this dorsal fiber pathway underpins both speech repetition and conceptually driven spoken language production, as underlying picture naming (e.g., Norman Geschwind, book Selected Papers on Language and the Brain, 1974).

Another proposal, however, holds that a ventral pathway underpinned by the uncinate fasciculus (UF) and extreme capsule (EmC) fiber tracts is primarily responsible for spoken language production, whereas the AF primarily underlies speech repetition (see the Lichtheim 2 model of Matthew Lambon Ralph and colleagues).

The results of computer simulations with WEAVER++/ARC revealed that the model accounts for the typical patterns of impaired and spared language performance associated with classic sudden-onset aphasias and one form of progressive aphasia (viz., semantic dementia). Moreover, the model accounts for evidence from patients with poststroke aphasia that damage to the AF but not the EmC/UF pathway predicts impaired production performance. These results demonstrate the viability of a dorsal-pathway account of language production.

Roelofs, A. (2014). A dorsal-pathway account of aphasic language production: The WEAVER++/ARC model. Cortex, 59, 33-48. Article (PDF 1339K)

 

 

WEAVER++/ARC on primary progressive aphasia

The WEAVER++/ARC model has not only been applied to poststroke aphasia, but also to aphasia resulting from neurodegenerative disease, called primary progressive aphasia (PPA). There are three major variants of PPA: nonfluent/agrammatic, semantic, and logopenic.

In a landmark article published in 1892, Arnold Pick (/pi:k/, in Czech pronounced as English peek, not as pick) argued that progressive diffuse atrophy of the brain may lead to focal disorders that resemble those due to stroke described by his contemporaries Wernicke and Lichtheim. The article reports on patient August H., who presented with progressive aphasia and circumscribed temporal lobe atrophy. To explain focal progressive disorders of language, Pick hypothesized "that simple progressive cerebral atrophy can also occasionally lead to the symptoms of a focal affection, perhaps through stronger local prominence of the diffuse process", or in the original German wording:

"dass die einfache progressive Hirnatrophie gelegentlich auch und zwar vielleicht durch stärkere locale Betonung des diffussen Processes zu den Symptomen einer Herdaffection führen kann" (Pick, 1892, p. 167).


Following this suggestion by Pick and modern empirical insights, WEAVER++/ARC assumes that PPA arises from a progressive loss of activation capacity in portions of the language network with neurocognitive epicenters specific to each PPA variant.

Computer simulations revealed that the model captures the patterns of impaired and spared naming, comprehension, and repetition performance of the PPA variants, at both group and individual patient levels. Moreover, it captures the worsening of performance with progression of the disease. The model explains about 90% of the variance, lending computational support to Pick’s suggestion and modern insights.

Roelofs, A. (2022). A neurocognitive computational account of word production, comprehension, and repetition in primary progressive aphasia. Brain and Language, 227, 105094. Article (PDF 2399K)

Pick, A. (1892). Ueber die Beziehungen der senilen Hirnatrophie zur Aphasie [On the relationships between senile brain atrophy and aphasia]. Prager Medicinische Wochenschrift, 17(16), 165-167.

 

Evidence indicates that ventral fiber tracts of the human brain play an important role in mediating visual-semantic information processing and its top-down control. The model explains how naming depends on ventral tract integrity in PPA.

 

Janssen, N., Roelofs, A., Mangnus, M., Sierpowska, J., Kessels, R. P. C., & Piai, V. (2020). How the speed of word finding depends on ventral tract integrity in primary progressive aphasia. NeuroImage: Clinical, 28, 102450. Article (PDF 3049K)

 

 

Clockwise from above: First page of the seminal 1892 article by Pick on August H., his clinic viewed from the outside and from the inside, and the apartment house in Prague where he lived.

 

In 1907, two articles appeared on the neuropathology of what later came to be called Alzheimer's disease. One article was by Alois Alzheimer working in the laboratory of Emil Kraepelin (a former postdoc of Wundt) in Munich and the other by Oskar Fischer working in the laboratory of Arnold Pick in Prague. Alzheimer described plaques and tangles in the brain of one patient (Auguste D.) and Fischer described the plaques in twelve patients. In 1911, Alzheimer reported on the neuropathology (i.e., spherical inclusion bodies) underlying the circumscribed temporal lobe atrophy described by Pick, later called Pick bodies. Andrew Kertesz and Pavel Kalvach (Archives of Neurology, 1996) give a historical account of Arnold Pick and German neuropsychiatry in Prague.

In a series of articles published between 1892 and 1904, Pick argued that a focal aphasia syndrome may arise from progressive atrophy of the left temporal lobe, sparing Wernicke's area. In 1913 (book Die agrammatischen Sprachstörungen), he posited that characteristics of agrammatic language production are the result of adaptation to the brain damage following an economy principle: With the remaining reduced capacity, telegraphic (i.e., simplified, elliptical) speech can still be produced, whereas full-fledged sentences are beyond capacity. This idea forms the basis of the modern adaptation theory of Herman Kolk, which provides the rationale for the SimpTell therapy app that aims to train elliptical style. Pick also advanced a stage theory of language production, foreshadowing the modern theories of Merrill Garrett and Pim Levelt that provide the general theoretical framework of the WEAVER++ model.

(Personal note A.R.: My work on progressive aphasia is dedicated to my parents, who both lost their language due to neurodegenerative disease)

 

WEAVER++/ARC on treatment effects in aphasia

Difficulties with word finding occasionally occur in all speakers, surfacing as pauses or errors, and commonly in all types of aphasia, often referred to as anomia. An understanding of anomia and its treatment-induced improvement is of fundamental scientific and clinical importance. Treatment employing semantic and phonological techniques has been successful in improving word finding and also functional communicative effectiveness. Phonological cueing involves the use of auditory cues that provide phonological information on the target word. For example, if a patient is unable to name a picture (e.g., of cat), spoken cues are given about the first sound (e.g., /k/) or the rhyme (e.g., /æt/) of the picture name, or the name itself is provided (e.g., /kæt/).

In the literature, disagreement exists about the locus of the cueing effects in aphasia and whether the effects arise from the same neurocognitive mechanisms as phonological effects in picture naming by healthy speakers. To expand our understanding of treatment effects, and to examine whether a unified account of the phonological effects in health and disease is possible, computer simulations with WEAVER were conducted. The model successfully simulates treatment effects in behavioural naming performance and neural measures. The simulations provide a proof of concept for the idea that treatment effects of phonological cueing in aphasia arise from the same neurocognitive mechanisms as immediate phonological effects in healthy speakers. This expands our understanding of word finding, associated difficulties, and their improvement by therapy.

Roelofs, A. (2021). Phonological cueing of word finding in aphasia: Insights from simulations of immediate and treatment effects. Aphasiology, 35, 169-185. Article (PDF 1451K)

 

WEAVER++ on reading and dyslexia

Since the seminal work of Martha Denckla and colleagues in the early 1970s (based on Norman Geschwind's hypothesis that color naming might predict reading), numerous studies have demonstrated that, in addition to phonological deficits, the majority of children with developmental reading disabilities also exhibit pronounced difficulties on naming-speed tasks, such as tests of "rapid automatized naming" (RAN). These tests require simple objects, colors, letters, or numbers to be named as quickly and accurately as possible. Naming speed is highly correlated with performance on word identification tasks, word reading efficiency measures, and measures of reading comprehension. Naming speed predicts later reading ability and helps identify risk at dyslexia in pre-literate children. Dyslexic readers are also known for their poor performance on Stroop color naming. Reading ability is negatively related to Stroop interference. Evidence suggest that attention mechanisms are critically implicated in reading and that disruption of these mechanisms may play a role in reading difficulties and dyslexia.

WEAVER++ provides functional analyses of object, color, and digit naming as well as word reading, and the model makes explicit how attention determines naming and reading. Moreover, the model provides an account of Stroop task performance and explains the negative linear relationship between reading skill and Stroop interference. It has been suggested that RAN is related to reading because reading recruits object-naming circuits in the left cerebral hemisphere. WEAVER++ makes explicit the connection between reading and object naming, both in functional and anatomical terms.

Roelofs, A. (2006). Functional architecture of naming dice, digits, and number words. Language and Cognitive Processes, 21, 78-111. Article (PDF 176K)

Protopapas, A., Archonti, A., & Skaloumbakas, C. (2007). Reading ability is negatively related to Stroop interference. Cognitive Psychology, 54, 251-282.  doi:10.1016/j.cogpsych.2006.07.003

 

 

 

WEAVER++ on specific language impairment

Evidence suggests that (subclinical) attention deficits also contribute to the impaired language performance of individuals with specific language impairment (SLI). This is a disorder of language acquisition and use in children who otherwise appear to be normally developing. The disorder may persist into adulthood. Difficulties concern language production (expressive language disorder) or both production and comprehension (mixed receptive-expressive language disorder). The features of the impaired language performance in SLI are quite variable, but common characteristics are a delay in starting to talk in childhood, deviant production of speech sounds, a restricted vocabulary, slow and inaccurate word retrieval (e.g. in picture naming), and use of simplified grammatical structures, including frequent omission of articles or plural and past tense endings (for a review, see Laurence Leonard’s 1998 book Children with Specific Language Impairment). In general, individuals with SLI seem to have a problem in dealing with relatively complex language structures, in both speech production and comprehension. A prominent account of SLI holds that these difficulties with complexity in language reflect a reduced capacity of systems underlying language processes, resulting from a limitation in general processing capacity (see the work of Laurence Leonard and colleagues). Moreover, it is becoming increasingly clear that attention deficits contribute to SLI.

Individuals with SLI appear to have reduced working memory capacity. Moreover, they may have deficits in sustained attention and attentional control. Capacity restrictions concerning language processes, working memory, and attention influence word planning in WEAVER++. For example, a capacity restriction in retrieving morphemes for a lemma may result in omission of inflectional morphemes, such as plural and past tense endings. This type of problem will be reinforced by capacity restrictions in working memory and attention. For word planning to be successful in the model, attention needs to be sustained until the phonological form has been planned and syllable motor programs may be accessed. Difficulties in maintaining attention will impede the planning process, especially when a complex mapping between levels is involved (e.g., such as the mapping between lemmas and morphemes).

Janssen, D. P., Roelofs, A., & Levelt, W.J.M. (2002). Inflectional frames in language production. Language and Cognitive Processes17, 209-236. Article (PDF 246K)

Levelt, W.J.M., Roelofs, A., & Meyer, A.S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22, 1-38. Article (PDF 693K)

Roelofs, A. (1996). Serial order in planning the production of successive morphemes of a word. Journal of Memory and Language, 35, 854-876. Article (PDF 249K)

Roelofs, A. (2006). Context effects of pictures and words in naming objects, reading words, and generating simple phrases. Quarterly Journal of Experimental Psychology, 59, 1764-1784. Article (PDF 172K)

Roelofs, A., & Piai, V. (2011). Attention demands of spoken word planning: A review. Frontiers in Psychology, 2, article 307. Article (PDF 976K)

 

WEAVER++ on Donders' seminal study and a modern replication

WEAVER++ has also been applied to the seminal RT data on speech repetition from Frans Donders and his students obtained in 1865 and from a modern replication (Roelofs, 2018).

Speech production RTs were first measured 150 years ago. On Monday, August 21, 1865, Donders conducted his seminal study with the a-, b-, and c-methods (i.e., simple, choice, go/no-go) using a speech repetition task, reported in his landmark article three years later. In 1874, in his Grundzüge der physiologischen Psychologie, Wundt criticized the c-method by arguing that it does involve a choice (i.e., whether or not to respond), whereas Donders maintained that it may not involve full discrimination.

A handwritten laboratory notebook of Donders was found to include unpublished data from him and his students. Analyses of these data revealed no b – c difference for his students, supporting Wundt's concern. Moreover, a modern replication of Donders' study using Donders' original stimulus lists yielded only a small b – c difference, supporting Wundt. Computer simulation with WEAVER++ of speech repetition under the a-, b-, and c-methods indicated that the difference between Donders and his students may plausibly result from choice in the c-method.

Roelofs, A. (2018). One hundred fifty years after Donders: Insights from unpublished data, a replication, and modeling of his reaction times. Acta Psychologica, 191, 228-233. Article (PDF 666K)

 

Wundt 2.0

Converging evidence for several of the assumptions implemented in WEAVER++ comes from simulations with the Wundt 2.0 model, which implements theoretical ideas from Wundt (1880, 1902/1904) about how attention controls naming. These assumptions shared between models include perceptual control throught distractor suppression and an architectural difference between naming and reading.

Roelofs, A. (2021). How attention controls naming: Lessons from Wundt 2.0. Journal of Experimental Psychology: General, 150, 1927-1955. Article (PDF 2577K)

Wundt, W. (1904). Principles of Physiological Psychology (E. B. Titchener, Trans.). Swan Sonnenschein. (Original work published 1902)

 

Orienting of attention in dual-task performance

Spatial orienting of attention may occur overtly or covertly, that is, with or without gaze shifting. If the stimuli for one or two tasks have different spatial positions, eye movements (gaze shifts) may need to occur between the stimuli. In an experimental paradigm used in my laboratory, speakers name stimuli (e.g., a picture of a cat with the word DOG superimposed) displayed on the left side of a computer screen and shift their gaze to an arrow (e.g., < or > flanked by two Xs) displayed on the right side of the screen to manually indicate its direction, see figure below. WEAVER++ describes how an attentional control process coordinates the multiple threads of processing in vocal responding, gaze shifting, and manual responding:

Roelofs, A. (2007). Attention and gaze control in picture naming, word reading, and word categorizing. Journal of Memory and Language, 57, 232-251. Article (PDF 311K)

Roelofs, A. (2008). Attention, gaze shifting, and dual-task interference from phonological encoding in spoken word planning. Journal of Experimental Psychology: Human Perception and Performance, 34, 1580-1598. Article (PDF 377K)

 

 

 

 

 

 

Planning stages

The model distinguishes between conceptual preparation, lemma retrieval, and word-form encoding, with the encoding of forms further divided into morphological, phonological, and phonetic encoding. During conceptual preparation, concepts are flagged as goal concepts. In lemma retrieval, a goal concept is used to retrieve a lemma from memory, which is a representation of the syntactic properties of a word, crucial for its use in sentences. For example, the lemma of the word cat says that it is a noun. Lemma retrieval makes these properties available for syntactic encoding processes. In word-form encoding, the lemma is used to retrieve the morpho-phonological properties of the word from memory in order to construct an appropriate articulatory program. For example, for cat the morpheme <cat> and the speech segments /k/, /æ/, and /t/ are retrieved and a phonetic plan for [kæt] is generated. Finally, articulation processes execute the motor program, which yields overt speech.

Assume a speaker wants to name a picture of a cat with the word DOG superimposed. This involves the conceptual identification of the picture based on the perceptual input and its designation as goal concept (i.e., CAT(X)), the retrieval of the lemma of the corresponding word (i.e., cat), and the encoding of the form of the word (i.e., [kæt]). The final result is a motor program for the word "cat", which can be articulated.

In performing the picture-word interference task, aspects of word planning require attention. First, attentional (executive, cognitive) control is needed. The system has to achieve picture naming rather than word reading ("goal control") and the irrelevant input (the word in picture naming) has to be suppressed ("input control"). In general, attentional control may engage the abilities of updating, inhibiting, and shifting (see the work of Akira Miyake and colleagues). Moreover, attentional control may regulate the sustaining of attention (alertness) and the orienting of attention, with or without shifting of eye gaze (as illustrated above). Attentional control is also needed for self-monitoring (i.e., monitoring internal or external speech), through which language users assess whether planning and performance are consistent with intent.

Real-world language use engages many components of language and cognition, and effective communication depends on interactions among several components. Thus, successful theories and models must treat all of these components in an integrated manner. The blueprint below (in Prussian blue) specifies how, according to the model, word planning relates to other cognitive abilities.

   

 

 

 

 

Neural substrates

Determining the neural substrates of word production components is not only of fundamental scientific importance but also of clinical relevance. For example, functional localization helps to understand poststroke and progressive aphasias and may assist in guiding brain surgeries that remove tumors or epileptic foci.

According to the model, word planning and attentional control are underpinned by extensive networks of brain areas. The figure below shows a lateral view of the left hemisphere of the human brain. Declarative memory underlying word planning is associated with a network of temporal and frontal areas, including Wernicke's and Broca's areas. Picture naming is achieved through picture perception, conceptual identification, lemma retrieval, word-form encoding, and articulatory processing. Reading aloud minimally involves word-form perception, word-form encoding, and articulatory processing. The attentional control system (indicated by rounded rectangles) is associated with lateral and medial frontal areas (including the anterior cingulate cortex, ACC, not shown) and parietal cortex. Procedural memory, containing the if-then rules, is associated with the basal ganglia, thalamus, frontal cortex (including Broca’s area), and cerebellum. The procedural system mediates between the declarative associative network and the attentional system. It is assumed that the declarative and procedural knowledge is stored cortically, while subcortical structures, like the hippocampus and basal ganglia, mediate the learning of the declarative and procedural knowledge, respectively.

Roelofs, A. (2014). A dorsal-pathway account of aphasic language production: The WEAVER++/ARC model. Cortex, 59, 33-48. Article (PDF 1339K) 

 

 

 

 

Much evidence suggests that the dorsolateral prefrontal cortex serves to maintain goals in working memory. ACC involvement in goal-referenced control agrees with the idea that attention is the principal link between cognition and motivation. For action control, it is not enough to have goals in working memory, but one should be motivated to attain them. Extensive projections from the thalamus and brainstem nuclei to the ACC suggest a role for drive and arousal. Extensive reciprocal connections between the ACC and dorsolateral prefrontal cortex suggest a role for working memory. The motor areas of the cingulate sulcus densely project to the brainstem, spinal cord, and motor cortex, which suggests a role of the ACC in motor control.

Roelofs, A., & Hagoort, P. (2002). Control of language use: Cognitive modeling of the hemodynamics of Stroop task performance. Cognitive Brain Research, 15, 85-97. Article (PDF 438K)

 

Homo Loquens: From monkey calls to human talking

We vocally communicate through speech but also with our cries and laughs. Whereas speech is learned, cries and laughs are innately specified. Such innate "calls" are observed in diverse vertebrates, including fish, amphibians, reptiles, birds, and mammals. Although the sound-producing organs differ (swim bladder in fish, syrinx in birds, larynx in amphibians, reptiles, and mammals), vertebrates seem to share a common brainstem and spinal cord organization for calls. Vocal production learning is common in birds (songbirds, parrots, and hummingbirds), but only humans, bats, cetaceans (dolphins, whales), and pinnipeds (seals) show evidence of vocal learning among mammals. At least in songbirds and humans, some of the forebrain pathways implicated in learned vocalization seem to share homologous components.

Both human and nonhuman primates (monkeys and apes) use their voice for communication. However, whereas an extensive network of perisylvian and medial cortical areas—including the ACC in some circumstances—is involved in the verbal vocal communication of humans (i.e., spoken word production), the only cortical area directly involved in call production by nonhuman primates is the ACC (see the work of Detlev Ploog, Uwe Jürgens, and colleagues). In nonhuman primates, the ACC plays a critical role in the voluntary initiation and suppression of calls (e.g., fear, alarm, aggression, and contact calls), which are all innate. The ACC also controls the innate vocalizations (e.g., crying, laughing, pain shrieking) of humans. The human ACC appears to be the cortical area where the evolutionary older innate-vocalization system and the newer spoken-word production system meet (see the work of Terrence Deacon). For a description of some commonalities of ACC function across call and word production, see

Roelofs, A. (2008). Attention to spoken word planning: Chronometric and neuroimaging evidence. Language and Linguistics Compass, 2, 389-405. Article (PDF 327K)

 

     

 

 

 

 

Declarative aspects

Declarative pieces of information ("facts") about words are stored in a labeled associative network, part of the brain's long-term declarative memory system. Labeled means that relationships between nodes are explicitly represented. Declarative memory has a relational organization (cf. Eichenbaum, 2012). The relational lexical network consists of three major strata: a conceptual stratum, a syntactic stratum, and a word-form stratum, corresponding to the major planning steps. The conceptual stratum represents conceptual facts as nodes and labeled links in a semantic network. For example, the concept CAT is represented by the node CAT(X) connected to ANIMAL(X) by an "is-a" link. The syntactic stratum contains lemma nodes, such as cat, which are connected by a "word class" link to nodes for their syntactic class (e.g., noun). Finally, the form stratum contains nodes representing morphemes (e.g., <cat>), segments (e.g., /k/), and motor programs (e.g., [kæt]). The figure above shows only a small fragment of the lexical network and most of the labels on the links have been omitted.

Another type of memory system that holds declarative information is working memory. In it, production goals, like the goal of naming a picture, are temporarily maintained.

 

 

 

 

Spreading activation

Information is retrieved from the associative declarative network by spreading activation. For example, a perceived entity (e.g., a cat) activates the corresponding concept node (i.e., CAT(X)) in the network. Activation then spreads through the network following a linear activation rule with a decay factor d. Each node m sends a proportion r of its activation to the nodes n it is connected to. For example, CAT(X) sends activation to other concepts such as ANIMAL(X) and DOG(X) and also to its lemma node cat.

 

WEAVER++'s lexical network is accessed by spreading activation while condition-action rules (see below) determine what is done with the activated lexical information depending on the goal. When a goal is placed in working memory, processing in the system is focused on those rules that include the goal among their conditions. The rules mediate attentional influences by selectively enhancing the activation of target nodes in the network in order to achieve mappings of targets onto articulatory programs. For example, in naming a picture of a cat, the activation of the concept node CAT(X) is selectively enhanced.

 

Attentional activation enhancements

The model assumes that the ACC is implicated in the attentional enhancement of the activation of targets (rather than performing conflict monitoring, as is often assumed, see the work of Jonathan Cohen, Matthew Botvinick, and colleagues). The attentional control system determines how strongly and for how long the enhancements occur, depending on the allocation policy (cf. Daniel Kahneman, Attention and Effort book, 1973; EPIC of David Meyer and David Kieras). Attention is assumed to be sustained to word planning just as long as is needed to achieve acceptable levels of speed and accuracy:

Roelofs, A., Van Turennout, M., & Coles, M. G. H. (2006). Anterior cingulate cortex activity can be independent of response conflict in Stroop-like tasks. Proceedings of the National Academy of Sciences USA, 103, 13884-13889. Article (PDF 350K)

Roelofs, A. (2008). Tracing attention and the activation flow in spoken word planning using eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34, 353-368. Article (PDF 285K)

Roelofs, A. (2008). Attention, gaze shifting, and dual-task interference from phonological encoding in spoken word planning. Journal of Experimental Psychology: Human Perception and Performance, 34, 1580-1598. Article (PDF 377K)

 

 

 

 

Procedural aspects

Procedural knowledge ("knowing how") is embodied by condition-action (if-then) rules, part of the brain's long-term procedural memory system. The "if-side" of a rule specifies a condition to be satisfied and the "then-side" of a rule specifies an action to be performed when the condition is met.

 

 

 

 

The model makes explicit how procedural knowledge interacts with declarative knowledge during word planning. It assumes "procedural attachment to nodes", as proposed in Artificial Intelligence by Terry Winograd (1975, Frame Representations and the Declarative/Procedural Controversy) and others.

Procedural knowledge is associated with, among others, the basal ganglia thalamocortical circuitry of the brain. This circuitry consists of a number of parallel loops that run from cortical areas (i.e., temporal, parietal, and frontal cortex, including Broca’s area) via the striatum (i.e., caudate nucleus and putamen) and the subthalamic nucleus to the globus pallidus and via the thalamus back to cortical areas, especially frontal cortex (including Broca’s area). The circuits enable the execution of an action (e.g., selecting a concept or a task set) whose condition (best) matches the current context and momentary goal. There is also a basal ganglia thalamocortical circuitry for the control of eye movements. The basal ganglia circuitry seems especially important for the learning and use of novel rather than expert procedural knowledge (see the work of Gregory Ashby and colleagues). Condition-action rules in procedural memory are taken to underlie both attentional control and more automatic processes in speaking. The procedural system is essential for translating declarative knowledge into action.  

 

 

 

 

Verification

Verification means that selections in human performance are accomplished by explicit reference to goals: goal-referenced control. The condition-action rules from procedural memory carry out the selection of nodes in declarative memory. A rule is triggered when its nodes become active. During word planning, a lemma retrieval rule selects a lemma if the connected concept is flagged as goal concept. For example, cat is selected for the concept CAT if it is the goal concept and cat has reached a critical difference in activation compared to other lemmas, as illustrated by the animation below.

 

 

The actual moment in time of firing of the rule is determined by the ratio of activation of the lemma node and the sum of all the others. Thus, how fast a node is selected depends on how active the other nodes are. In short, there is activation-based triggering and firing of condition-action rules.

 

 

 

 

A morphological rule selects the morpheme nodes that are connected to the selected lemma and its morphosyntactic parameter values, such as singular or plural (the stem <cat> is selected for cat + singular). Phonological rules select the segments that are connected to the selected morphemes (/k/, /æ/, and /t/ for <cat>) and syllabify the segments (e.g., /k/ is made syllable onset: onset(/k/)) to create a phonological word representation. Finally, phonetic rules select syllable-based articulatory programs that are appropriately connected to the syllabified segments (i.e., [kæt] is selected for onset(/k/), nucleus(/æ/), and coda(/t/)). The actual moment of selection of syllable program nodes is also determined by a ratio of activations, such that how fast selection occurs depends on how active other nodes are.

 

From Wilhelm Wundt via Watt at Würzburg to WEAVER++

 

Although issues concerning the attentional control of human performance were explored in the early days of experimental psychology by Frans Donders, James McKeen Cattell, and Wilhelm Wundt (mental chronometry), no real progress was made in understanding the mechanisms of control (but see Wundt, 1880, 1902/1904). Associationist and behaviorist theoreticians, from David Hume to Burrhus Skinner, accounted for action selection by postulating associations between stimuli and responses. However, if all our actions were determined exclusively by stimulus-response associations, goals could not determine which action to make because the strongest association would automatically determine the response. Around 1900, the Würzburg school with Narziss Ach, Oswald Külpe, and Henry Watt demonstrated the importance of the task ("Aufgabe") in determining a response. However, how exactly task goals directed processing remained unclear. In the 1910s, Georg Müller proposed an account in associative terms, whereas Otto Selz proposed an account in terms of symbolic structures and rules. Later theoretical developments are descendants of these ideas. On the view that is still prominent in the attention and performance literature, goals associatively bias or "sculpt" the activation of one response pathway (e.g., for picture naming, in responding to the picture of a cat with the word DOG superimposed) rather than another (e.g., for oral reading), following Müller. On another view, following Selz, and computationally implemented in WEAVER++, attentional control arises from explicit reference to goals, accomplished by condition-action rules. Pictures below (from left to right): Henry Watt, Oswald Külpe, Otto Selz, and Allen Newell and Herbert Simon.

 

 

 

The idea of goal-referenced control that originated with Otto Selz in the 1910s flourished in the work of Adriaan de Groot, Allen Newell and Herbert Simon, and John Anderson, among others, on higher-level cognitive processes like problem solving (e.g., playing chess, proving logic theorems, and solving puzzles such as the Tower of Hanoi), where associative models generally failed. However, due to the traditional partitioning of experimental psychology into cognition, perception, and action, with little communication across the boundaries, the idea of goal-referenced control long has had little impact in the perception-action literature. However, in the past few decades, goal-referenced control has made successful strides into the attention and performance literature.

That goal-referenced control underlies both problem-solving and performing Stroop-like tasks agrees with the strong connection between attentional control and general intelligence (Charles Spearman's g, see the work of John Duncan and colleagues, reviewed by Duncan in his 2010 book How Intelligence Happens). Solving intelligence-demanding problems, like the Raven Matrices or the Tower of Hanoi puzzle, involves working serially from one subgoal to another, each with focused attention, until the overall goal is achieved. Individual differences in general intelligence are most pronounced in behavioral measures when attentional control is required (see the work of Randall Engle and colleagues).

In his dissertation work at Würzburg University, Henry Watt found that when verbal responses of the same intrinsic speed were grouped together, a variation of task had a similar effect across latency groups, although he did not quantify this effect. He stated, "The influence of the task is independent of the rapidity of the tendency to reproduction itself" (Watt, 1906, Journal of Anatomy and Physiology, p. 260, original italics). Watt's regularity has recently been confirmed and quantified using modern techniques for analysing response time distributions:

Roelofs, A. (2008). Dynamics of the attentional control of word retrieval: Analyses of response time distributions. Journal of Experimental Psychology: General, 137, 303-323. Article (PDF 392K)

 

 

 

 

Planning latencies

Given the equations for spreading activation and rule firing, the mathematically expected mean planning latencies can be computed.

 

 

The model accounts not only for mean response times but also for characteristics of the shape of response time distributions, including measures related to variability and skewness (e.g., as assessed by ex-Gaussian and delta-plot analyses).

San José, A., Roelofs, A., & Meyer, A. S. (2021). Modeling the distributional dynamics of attention and semantic interference in word production. Cognition. Article (PDF 4121K)

 

 

 

 

Frequently asked questions (FAQs)

 

Question: Why the name WEAVER?

Answer: Apart from providing a useful acronym (Word Encoding by Activation and VERification) and metaphor, it is accociated with an interesting (hi)story.

Weaving is a psychomotor skill performed already in very early times, known in the prehistoric era well before the ancient Egyptian and Greek civilizations. Weaving plays a prominent role in several Greek myths, including those featuring Arachne, Penelope, and Philomela. Arachne engaged in a weaving contest with the weaver goddess Athena, who transformed her into a spider, destined to weave forever. Penelope wove her design for a shroud by day but unraveled it again at night to keep her suitors from claiming her while waiting for Odysseus to return from the Trojan war. And Philomela, whose tongue was cut out after she was raped, used a loom as her voice to tell about her violation in a woven design. The link between weaving and communication is evident from the English word text, which is derived from the Latin word for weaving, texare. More recently, in the early nineteenth century, Jacquard’s invention of a programmable loom, which used punch cards with stored instructions for weaving patterns, led to the development of the modern computer (for a magnificent account, see James Essinger’s Jacquard’s web: How a Hand-Loom Led to the Birth of the Information Age, 2007). As Ada Lovelace (1843) stated about the Analytical Engine (the first proposed general-purpose computer) of Charles Babbage: "The Analytical Engine weaves algebraic patterns just as the Jacquard loom weaves flowers and leaves".

It is important to note, however, that the "weaving" of words during speaking differs in a fundamental respect from most other human psychomotor skills. As Charles Darwin stated, "Man has an instinctive tendency to speak, as we see in the babble of our young children, whereas no child has an instinctive tendency to bake, brew, or write."

 

Question: Why computational modeling? Aren't intuitions or verbal verbal theories just as good?

Answer: This question is best answered by quoting Daniel Mirman and Allison Britt (2014, Philos Trans R Soc Lond), who stated that "intuitions and verbal theories are not enough because they can be claimed to predict (or not predict) just about anything. Computational models provide a concrete implementation of a proposed theory that can then be tested empirically to evaluate whether it truly accounts for the observed data" (p. 11).

 

Question: Why do we need labeled links?

Answer: A mere associative link between two nodes tells nothing about the relation between the entities represented. For example, the concept CAT(X) is strongly associated with both DOG(X) and ANIMAL(X) but the relationship between CAT(X) and DOG(X) is very different from the relationship between CAT(X) and ANIMAL(X). The importance of explicitly representing the relation between nodes was recognized by Otto Selz in the early 1900s, and labeled link have become a central part of semantic networks in Artificial Intelligence since the seminal work of Ross Quillian in the late 1960s. Imagine the Internet without labeled (hyper)links! Reseach in cognitive neuroscience has characterized declaractive memory as having a relational organization (see The Cognitive Neuroscience of Memory: An Introduction by Howard Eichenbaum, 2012).

 

Question: Why do we need condition-action rules? Wouldn't it be better to have a purely associative model?

Answer: There exists good evidence that language performance involves both declarative aspects (structured symbolic representations) and procedural aspects (if-then production rules), underpinned by the brain's declarative and procedural systems (e.g., the work of Michael Ullman and colleagues). Evidence suggests that condition-action rules underlie both the linguistic processes associated with the word planning stages (i.e., conceptual identification, lemma retrieval, word-form encoding) as well as executive control processes, associated with frontoparietal and basal ganglia thalamocortical networks in the brain (e.g., the work of Earl Miller and colleagues and John Duncan and colleagues).

For a theoretical account of how structured symbolic representations and if-then production rules may be realized by networks of spiking neurons, see the work of Chris Eliasmith and colleagues (see Eliasmith's 2013 book How to Build a Brain).



Question: Doesn't the evidence that word planning requires attention violate the assumption that the planning components are autonomous?

Answer: The term "autonomous" refers to at least two properties, which may hold independently: automaticity and informational encapsulation. That word planning components require some attention means that they do not run completely on their own resources. Processing components are informationally encapsulated if they operate with restricted knowledge only, mapping specific input onto specific output. For example, the phonological encoding component maps one or more morphemes onto a phonological word representation, requiring morpho-phonological knowledge but no other information, like semantics. Other researchers do not assume such informational encapsulation (e.g., Caramazza and colleagues assume that the management of the articulatory buffer uses semantic information). A processing component may be informationally encapsulated while not operating fully automatically, as is the case in WEAVER++. Similarly, Jerry Fodor (1983, The Modularity of Mind) assumes that "modules" are informationally encapsulated but presumably do not run fully automatically.



Question: Where can I find information on WEAVER++?

Answer: Here are a number of references:

Roelofs, A. (2022). A neurocognitive computational account of word production, comprehension, and repetition in primary progressive aphasia. Brain and Language, 227, 105094. Article (PDF 2399K)

Roelofs, A. (2018). A unified computational account of cumulative semantic, semantic blocking, and semantic distractor effects in picture naming. Cognition, 172, 59-72. Article (PDF 574K)

Roelofs, A. (2014). A dorsal-pathway account of aphasic language production: The WEAVER++/ARC model. Cortex, 59, 33-48. Article (PDF 1339K)

Roelofs, A. (2008). Dynamics of the attentional control of word retrieval: Analyses of response time distributions. Journal of Experimental Psychology: General, 137, 303-323. Article (PDF 392K)

Roelofs, A., Van Turennout, M., & Coles, M. G. H. (2006). Anterior cingulate cortex activity can be independent of response conflict in Stroop-like tasks. Proceedings of the National Academy of Sciences USA, 103, 13884-13889. Article (PDF 350K)

Roelofs, A. (2003). Goal-referenced selection of verbal action: Modeling attentional control in the Stroop task. Psychological Review, 110, 88-125. Article (PDF 585K)

Roelofs, A., & Hagoort, P. (2002). Control of language use: Cognitive modeling of the hemodynamics of Stroop task performance. Cognitive Brain Research, 15, 85-97. Article (PDF 438K)

A complete list of articles that report on WEAVER++ simulations can be found here.

 

 

Question: What programming language has been used for the model, and is the program available?

Answer: Most of the WEAVER++ and WEAVER++/ARC applications have been programmed in the C programming language using the Microsoft Visual C++ environment. These programs are available from Ardi Roelofs or from the Open Science Framework (OSF) at https://osf.io/ue4bn/. Applications in Python have been made by Aitor San José.

 

Track record of WEAVER++ (WEAVER++'s Web)