Similarity-induced interference or facilitation in language production reflects representation, not selection.

Researchers have long interpreted the presence or absence of semantic interference in picture naming latencies as confirming or refuting theoretical claims regarding competitive lexical selection. But inconsistent empirical results challenge any mechanistic interpretation. A behavioral experiment first verified an apparent boundary condition in a blocked picture naming task: when orthogonally manipulating association type, taxonomic associations consistently elicit interference, while thematic associations do not. A plausible representational difference is that thematic feature activations depend more on supporting contexts. Simulations show that context-sensitivity emerges from the distributional statistics that are often used to measure thematic associations: residual semantic activation facilitates the retrieval of words that share semantic features, counteracting learning-based interference, and training a production model with greater sequential cooccurrence for thematically related words causes it to acquire stronger residual activation for thematic features. Modulating residual activation, either directly or through training, allows the model to capture gradient values of interference and facilitation, and in every simulation competitive and noncompetitive selection algorithms produce qualitatively equivalent results.

Rapid learning of a phonemic discrimination in the first hours of life.

Human neonates can discriminate phonemes, but the neural mechanism underlying this ability is poorly understood. Here we show that the neonatal brain can learn to discriminate natural vowels from backward vowels, a contrast unlikely to have been learnt in the womb. Using functional near-infrared spectroscopy, we examined the neuroplastic changes caused by 5 h of postnatal exposure to random sequences of natural and reversed (backward) vowels (T1), and again 2 h later (T2). Neonates in the experimental group were trained with the same stimuli as those used at T1 and T2. Compared with controls, infants in the experimental group showed shorter haemodynamic response latencies for forward vs backward vowels at T1, maximally over the inferior frontal region. At T2, neural activity differentially increased, maximally over superior temporal regions and the left inferior parietal region. Neonates thus exhibit ultra-fast tuning to natural phonemes in the first hours after birth.

The psychological reality of picture name agreement.

Picture name agreement is commonly used as both a control variable and independent variable in studies of language production. It describes the proportion of participants who volunteer a picture's modal name in a norming study-a population-level descriptor-but researchers often assume that name agreement also indexes cognitive processes that occur within individuals. For instance, if norms show that 50% of speakers name a picture as couch, then each time a person tries to name the picture, they might have a 50% chance of selecting couch. An alternative, however, is that name agreement may simply reflect population-level sampling of more stable individual preferences (e.g., 50% of speakers prefer the name couch), continually developed through experience. One way to distinguish between these possibilities - and assess the psychological reality of name agreement - is simply to re-norm pictures with the same individuals. In Experiment 1, we therefore collected timed naming norms for a large set of line drawings from the same 25 native British English speakers twice, 1-2 weeks apart. Results show participants' name choices in Session 2 are jointly predicted by population-level name agreement, from our previous norms, and individuals' own productions in Session 1. Experiment 2 replicated this result and further showed that prior selections predicted Session 3 outcomes better than those in Session 2, in line with an incremental learning account. This is the first direct demonstration that picture name agreement has some psychological validity, but also reveals that it does not directly index within-participant lexical competition as previously assumed.

Audiovisual Learning in Dyslexic and Typical Adults: Modulating Influences of Location and Context Consistency.

Learning to read involves efficient binding of visual to auditory information. Aberrant cross-modal binding skill has been observed in both children and adults with developmental dyslexia. Here, we examine the contribution of episodic memory to acquisition of novel cross-modal bindings in typical and dyslexic adult readers. Participants gradually learned arbitrary associations between unfamiliar Mandarin Chinese characters and English-like pseudowords over multiple exposures, simulating the early stages of letter-to-letter sound mapping. The novel cross-modal bindings were presented in consistent or varied locations (i.e., screen positions), and within consistent or varied contexts (i.e., co-occurring distractor items). Our goal was to examine the contribution, if any, of these episodic memory cues (i.e., the contextual and spatial properties of the stimuli) to binding acquisition, and investigate the extent to which readers with and without dyslexia would differ in their reliance on episodic memory during the learning process. Participants were tested on their ability to recognize and recall the bindings both during training and then in post-training tasks. We tracked participants' eye movements remotely with their personal webcams to assess whether they would re-fixate relevant empty screen locations upon hearing an auditory cue-indicative of episodic memory retrieval-and the extent to which the so-called "looking-at-nothing behavior" would modulate recognition of the novel bindings. Readers with dyslexia both recognized and recalled significantly fewer bindings than typical readers, providing further evidence of their persistent difficulties with cross-modal binding. Looking-at-nothing behavior was generally associated with higher recognition error rates for both groups, a pattern that was particularly more evident in later blocks for bindings encoded in the inconsistent location condition. Our findings also show that whilst readers with and without dyslexia are capable of using stimulus consistencies in the input-both location and context-to assist in audiovisual learning, readers with dyslexia appear particularly reliant on consistent contextual information. Taken together, our results suggest that whilst readers with dyslexia fail to efficiently learn audiovisual binding as a function of stimulus frequency, they are able to use stimulus consistency-aided by episodic recall-to assist in the learning process.

"Did I Say Cherry?" Error Patterns on a Blocked Cyclic Naming Task for Bilingual Children With and Without Developmental Language Disorder.

Purpose Using a blocked cyclic picture-naming task, we compared accuracy and error patterns across languages for Spanish-English bilingual children with and without developmental language disorder (DLD). Method Pictured stimuli were manipulated for semantic similarity across two (Same and Mixed) category contexts. Children's productions were scored off-line for accuracy, error frequency, and error type. Results Typically developing children were more accurate and produced fewer errors than their peers with DLD; however, this was moderated by task language and semantic context. Children were generally more accurate when naming pictures in English compared to Spanish and in the Mixed-category context compared to the Same-category context. Analyses of error types further showed that children with less English language exposure specifically produced more nonresponses in English than in Spanish. Children with DLD produced more of each error type than their typically developing peers, particularly in Spanish. Conclusions Regardless of language ability, bilingual children demonstrated greater difficulty with lexical retrieval for pictured items in the semantically related context than in the unrelated context. However, bilingual children with DLD produced more errors of all types than is typical for children developing more than one language. Their greater error rates are not secondary to limited second language exposure but instead reflect deficits inherent to the nature of language impairment. Results from this study are discussed using a framework of semantic constraint, where we propose that because bilingual children with DLD have impoverished semantic networks, and this knowledge insufficiently constrains activation for lexical selection, thereby increasing error production.

Longitudinal evidence for simultaneous bilingual language development with shifting language dominance, and how to explain it.

Theories of how language works have shifted from rule-like competence accounts to more skill-like incremental learning accounts. Under these, people acquire language incrementally, through practice, and may even lose it incrementally as they acquire competing mappings. Incremental learning implies that (1) a bilingual's abilities in their languages should depend on how much they practice each (not merely age of acquisition), and (2) using an L2 more could cause a bilingual to gradually 'unlearn' their L1. Using timed picture naming and vocabulary measures, we tracked 139 children for several years as they transitioned from mostly-Spanish homes to mostly-English schools. Following their increased English use, many became more proficient in English than Spanish around the third grade, demonstrating continual learning. But their Spanish also improved, showing that L1-attrition is not inevitable. Incremental learning explains both co-improvement and L1-attrition as consequences of experience-driven learning: improvement from continuing L1 use can offset competitive unlearning.

Bilinguals apply language-specific grain sizes during sentence reading.

Languages differ in the consistency with which they map orthography to phonology, and a large body of work now shows that orthographic consistency determines the style of word decoding in monolinguals. Here, we characterise word decoding in bilinguals whose two languages differ in orthographic consistency, assessing whether they maintain two distinct reading styles or settle on a single 'compromise' reading style. In Experiment 1, Welsh-English bilinguals read cognates and pseudowords embedded in Welsh and English sentences. Eye-movements revealed that bilinguals dynamically alter their decoding strategy according to the language context, including more fixations during lexical access for cognates in the more consistent orthography (Welsh) than in the less consistent orthography (English), and these effects were specific to word (as opposed to pseudoword) processing. In Experiment 2, we compared the same bilinguals' eye movements in the English sentence reading context to those of monolinguals'. Bilinguals' eye-movement behaviour was very similar to monolinguals' when reading English, suggesting that their knowledge of the more consistent orthography (Welsh) did not alter their decoding style when reading in English. This study presents the first characterisation of bilingual decoding style in sentence reading. We discuss our findings in relation to connectionist reading models and models of bilingual visual word recognition.

The paca that roared: Immediate cumulative semantic interference among newly acquired words.

With 40,000 words in the average vocabulary, how can speakers find the specific words that they want so quickly and easily? Cumulative semantic interference in language production provides a clue: when naming a large series of pictures, with a few mammals sprinkled about, naming each subsequent mammal becomes slower and more error-prone. Such interference mirrors predictions from an incremental learning algorithm applied to meaning-driven retrieval from an established vocabulary, suggesting retrieval benefits from a constant, implicit, re-optimization process (Oppenheim et al., 2010). But how quickly would a new mammal (e.g. paca) engage in this re-optimization? In this experiment, 18 participants studied 3 novel and 3 familiar exemplars from each of six semantic categories, and immediately performed a timed picture-naming task. Consistent with the learning model's predictions, naming latencies revealed immediate cumulative semantic interference in all directions: from new words to new words, from new words to old words, from old words to new words, and from old words to old words. Repeating the procedure several days later produced similar-magnitude effects, demonstrating that newly acquired words can be immediately semantically integrated, at least to the extent necessary to produce typical cumulative semantic interference. These findings extend the Dark Side model's scope to include novel word production, and are considered in terms of mechanisms for lexical selection.

Interactions between Lexical Access and Articulation.

This study investigates the interaction of lexical access and articulation in spoken word production, examining two dimensions along which theories vary. First, does articulatory variation reflect a fixed plan, or do lexical access-articulatory interactions continue after response initiation? Second, to what extent are interactive mechanisms hard-wired properties of the production system, as opposed to flexible? In two picture-naming experiments, we used semantic neighbor manipulations to induce lexical and conceptual co-activation. Our results provide evidence for multiple sources of interaction, both before and after response initiation. While interactive effects can vary across participants, we do not find strong evidence of variation of effects within individuals, suggesting that these interactions are relatively fixed features of each individual's production system.

A blind spot in correct naming latency analyses.

Speech errors and naming latencies provide two complementary sets of behavioural data for understanding language production processes. A recent analytical trend-applied to intact and impaired production alike-highlights a link between specific features of correct picture naming latency distributions and the retrieval processes thought to underlie them. Although chronometric approaches to language production typically consider correct response times in isolation, adequately accounting for their distributions in error-prone situations requires also considering the errors that sometimes censor them. In this paper, I illustrate by simulation how excluding incorrect word retrievals predictably alters observed distributions of correct naming latencies. To the extent that naming errors impose a stochastic deadline on successful production, their censoring should tend to reduce the mean, variance, and skew of observed latencies for correct responses, relative to the uncensored underlying distribution.

Inner speech as a forward model?

Pickering & Garrod (P&G) consider the possibility that inner speech might be a product of forward production models. Here I consider the idea of inner speech as a forward model in light of empirical work from the past few decades, concluding that, while forward models could contribute to it, inner speech nonetheless requires activity from the implementers.

The case for subphonemic attenuation in inner speech: comment on Corley, Brocklehurst, and Moat (2011).

Corley, Brocklehurst, and Moat (2011) recently demonstrated a phonemic similarity effect for phonological errors in inner speech, claiming that it contradicted Oppenheim and Dell's (2008) characterization of inner speech as lacking subphonemic detail (e.g., features). However, finding an effect in both inner and overt speech is not the same as finding equal effects in inner and overt speech. In this response, I demonstrate that Corley et al.'s data are entirely consistent with the notion that inner speech lacks subphonemic detail and that each of their experiments exhibits a Similarity × Articulation interaction of about the same size that Oppenheim and Dell (2008, 2010) reported in their work. I further show that the major discrepancy between the labs' data lies primarily in the magnitude of the main effect of phonemic similarity and the overall efficiency of error elicitation, and demonstrate that greater similarity effects are associated with lower error rates. This leads to the conclusion that successful speech error research requires finding a sweet spot between too much randomness and not enough data.

Motor movement matters: the flexible abstractness of inner speech.

Inner speech is typically characterized as either the activation of abstract linguistic representations or a detailed articulatory simulation that lacks only the production of sound. We present a study of the speech errors that occur during the inner recitation of tongue-twister-like phrases. Two forms of inner speech were tested: inner speech without articulatory movements and articulated (mouthed) inner speech. Although mouthing one's inner speech could reasonably be assumed to require more articulatory planning, prominent theories assume that such planning should not affect the experience of inner speech and, consequently, the errors that are "heard" during its production. The errors occurring in articulated inner speech exhibited the phonemic similarity effect and the lexical bias effect--two speech-error phenomena that, in overt speech, have been localized to an articulatory-feature-processing level and a lexical-phonological level, respectively. In contrast, errors in unarticulated inner speech did not exhibit the phonemic similarity effect--just the lexical bias effect. The results are interpreted as support for a flexible abstraction account of inner speech. This conclusion has ramifications for the embodiment of language and speech and for the theories of speech production.

The dark side of incremental learning: a model of cumulative semantic interference during lexical access in speech production.

Naming a picture of a dog primes the subsequent naming of a picture of a dog (repetition priming) and interferes with the subsequent naming of a picture of a cat (semantic interference). Behavioral studies suggest that these effects derive from persistent changes in the way that words are activated and selected for production, and some have claimed that the findings are only understandable by positing a competitive mechanism for lexical selection. We present a simple model of lexical retrieval in speech production that applies error-driven learning to its lexical activation network. This model naturally produces repetition priming and semantic interference effects. It predicts the major findings from several published experiments, demonstrating that these effects may arise from incremental learning. Furthermore, analysis of the model suggests that competition during lexical selection is not necessary for semantic interference if the learning process is itself competitive.

Saying the right word at the right time: Syntagmatic and paradigmatic interference in sentence production.

Retrieving a word in a sentence requires speakers to overcome syntagmatic, as well as paradigmatic interference. When accessing cat in "The cat chased the string," not only are similar competitors such as dog and cap activated, but also other words in the planned sentence, such as chase and string. We hypothesize that both types of interference impact the same stage of lexical access, and review connectionist models of production that use an error-driven learning algorithm to overcome that interference. This learning algorithm creates a mechanism that limits syntagmatic interference, the syntactic "traffic cop," a configuration of excitatory and inhibitory connections from syntactic-sequential states to lexical units. We relate the models to word and sentence production data, from both normal and aphasic speakers.

Inner speech slips exhibit lexical bias, but not the phonemic similarity effect.

Inner speech, that little voice that people often hear inside their heads while thinking, is a form of mental imagery. The properties of inner speech errors can be used to investigate the nature of inner speech, just as overt slips are informative about overt speech production. Overt slips tend to create words (lexical bias) and involve similar exchanging phonemes (phonemic similarity effect). We examined these effects in inner and overt speech via a tongue-twister recitation task. While lexical bias was present in both inner and overt speech errors, the phonemic similarity effect was evident only for overt errors, producing a significant overtness by similarity interaction. We propose that inner speech is impoverished at lower (featural) levels, but robust at higher (phonemic) levels.