Perceptual Bias in Speech Error Data Collection: Insights from Spanish Speech Errors

This paper studies the reliability and validity of naturalistic speech errors as a tool for language production research. Possible biases when collecting naturalistic speech errors are identified and specific predictions derived. These patterns are then contrasted with published reports from Germanic languages (English, German and Dutch) and one Romance language (Spanish). Unlike findings in the Germanic languages, Spanish speech errors show many patterns which run contrary to those expected from bias: (1) more phonological errors occur between words than within word; (2) word-initial consonants are less likely to participate in errors than word-medial consonants, (3) errors are equally likely in stressed and in unstressed syllables, (4) perseverations are more frequent than anticipations, and (5) there is no trace of a lexical bias. We present a new corpus of Spanish speech errors collected by many theoretically naïve observers (whereas the only corpus available so far was collected by two highly trained theoretically informed observers), give a general overview of it, and use it to replicate previous reports. In spite of the different susceptibility of these methods to bias, results were remarkably similar in both corpora and again contrary to predictions from bias. As a result, collecting speech errors “in the wild” seems to be free of bias to a reasonable extent even when using a multiple-collector method. The observed contrasting patterns between Spanish and Germanic languages arise as true cross-linguistic differences.     

The lexical bias effect in bilingual speech production: Evidence for feedback between lexical and sublexical levels across languages

The lexical bias effect (LBE) is the tendency for phonological substitution errors to result in existing words (rather than nonwords) at a rate higher than would be predicted by chance. This effect is often interpreted as revealing feedback between the phonological and lexical levels of representation during speech production. We report two experiments in which we tested for the LBE (1) in second-language production (Experiment 1) and (2) across the two languages of a bilingual (Experiment 2). There was an LBE in both situations. Thus, to the extent that the LBE reveals the presence of interactivity between the phonological and the lexical levels of representation, these effects suggest that there is feedback in second-language production and that it extends across the two languages of a bilingual.     

On the autonomy of phonological encoding: Evidence from slips of the tongue in Spanish

The present study focuses on the relationship between the processes of lexical retrieval and phonological encoding in sentence production. An analysis of spontaneous slips of the tongue in Spanish reveals that (1) there is as yet no clear evidence for a lexical bias effect on sublexical errors (segment movement and substitution), and hence for positive feedback from phonological encoding to the level of lexical representation; and (2) meaning and form appear to be largely dissociated in lexical errors, which lends support to the hypothesis that lexical retrieval proceeds in two independent stages during sentence production. These findings are discussed in the light of two alternative accounts of language production mechanisms: the interactive activation model and the structural, processing-stages model.This research has been supported by grant no. PB87-0531 from the Dirección General de Investigación Científica y Técnica, Spanish Ministry of Education. We thank Kathryn Bock, Gary Dell, and David Swinney for helpful comments and discussion.     

The Influence of Phonological Context on the Sound Errors of a Speaker with Wernicke's Aphasia

A corpus of phonological errors produced in narrative speech by a Wernicke's aphasic speaker (R.W.B.) was tested for context effects using two new methods for establishing chance baselines. A reliable anticipatory effect was found using the second method, which estimated chance from the distance between phoneme repeats in the speech sample containing the errors. Relative to this baseline, error-source distances were shorter than expected for anticipations, but not perseverations. R.W.B.'s anticipation/perseveration ratio measured intermediate between a nonaphasic error corpus and that of a more severe aphasic speaker (both reported in Schwartz et al., 1994), supporting the view that the anticipatory bias correlates to severity. Finally, R.W.B's anticipations favored word-initial segments, although errors and sources did not consistently share word or syllable position.     

Evidence for an activation locus of the word-frequency effect in lexical decision

The authors report a lexical decision experiment designed to determine whether activation is the locus of the word-frequency effect. K. R. Paap and L. S. Johansen (1994) reported that word frequency did not affect lexical decisions when exposure durations were brief; they accounted for this by proposing that data-limited conditions prevented late-occurring verification processes. Subsequently, P. A. Allen, A. F. Smith, M. Lien, T. A. Weber, and D. J. Madden (1997) and K. R. Paap, L. S. Johansen, E. Chun, and P. Vonnahme (2000) reported additional evidence that word-frequency effects do and do not have an activation locus, respectively. The authors further tested this issue in a lexical decision experiment using data-limited procedures--predicted by verification models to eliminate word-frequency effects. The authors observed word-frequency effects using individually determined exposure durations that were only 1 screen cycle longer than the exposure duration that yielded chance performance. Word-frequency effects persisted even when an adjusted measure of performance was used.     

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.     

Reading strategies in English by adults whose first language is Spanish

The performance of nine Spanish speakers on tests of single-word reading and phonological awareness in English was examined and compared to that of monolingual adults and to that of monolingual children of similar reading ability. Even though the Spanish participants had several years experience of reading and writing in English and performed well at reading nonwords, they showed little evidence of phonological processing strategies when reading familiar words. For example, they performed relatively poorly at written rhyme judgements despite good performance on tests of phonological awareness. When compared to monolingual English children, the Spanish speakers made fewer phonological errors on tests of visual lexical decision and written homophone definition. Unlike the children, they showed no evidence of a regularity effect in reading. Only on unfamiliar words did they show evidence of phonological processing in reading. Possible explanations of this pattern of performance are discussed.     

Language identification from visual-only speech signals

Our goal in the present study was to examine how observers identify English and Spanish from visual-only displays of speech. First, we replicated the recent findings of Soto-Faraco et al. (2007) with Spanish and English bilingual and monolingual observers using different languages and a different experimental paradigm (identification). We found that prior linguistic experience affected response bias but not sensitivity (Experiment 1). In two additional experiments, we investigated the visual cues that observers use to complete the languageidentification task. The results of Experiment 2 indicate that some lexical information is available in the visual signal but that it is limited. Acoustic analyses confirmed that our Spanish and English stimuli differed acoustically with respect to linguistic rhythmic categories. In Experiment 3, we tested whether this rhythmic difference could be used by observers to identify the language when the visual stimuli is temporally reversed, thereby eliminating lexical information but retaining rhythmic differences. The participants performed above chance even in the backward condition, suggesting that the rhythmic differences between the two languages may aid language identification in visual-only speech signals. The results of Experiments 3A and 3B also confirm previous findings that increased stimulus length facilitates language identification. Taken together, the results of these three experiments replicate earlier findings and also show that prior linguistic experience, lexical information, rhythmic structure, and utterance length influence visual-only language identification.     

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.     

Comprehension of passives in Broca's aphasia

Drai and Grodzinsky have statistically analyzed a large corpus of data on the comprehension of passives by patients with Broca's aphasia. The data come, according to Drai and Grodzinsky, from binary choice tasks. Among the languages that are analyzed are Dutch and German. Drai and Grodzinsky argue that Dutch and German speaking Broca patients should be relatively good (that is, perform above chance) on comprehension of passive sentences, since in Dutch and German passives the relative order of the object and lexical verb is the underlying order and hence no movement takes place. We will demonstrate that both their linguistic arguments and their selection of Dutch data are invalid.     

Acquired dyslexia in Spanish: a review and some observations on a new case of deep dyslexia

Readers and writers of Spanish use an orthography that is highly transparent. It has been proposed that readers of Spanish can rely on grapheme-phoneme correspondences, alone, to access meaning or phonology from print. In recent years, a number of case studies have yielded evidence inconsistent with this idea. We review these studies with particular focus on those that report evidence for reading based on direct lexical mappings between print, orthographic representations, and meaning or phonology. We report a new case of acquired literacy impairment in Spanish, MJ, who presents a pattern of preserved abilities and deficits symptomatic of deep dyslexia. The patient is unable to read nonwords, but can read a substantial number of words. Her reading is characterized by the production of semantic, visual, and derivational errors. We argue that MJ has a deficit in her lexical selection ability, common to both her reading and her naming problems. We propose that MJ, and the other cases we review, demonstrate that lexical reading is adopted by skilled readers even in a transparent language.     

Formal lexical paragraphias in a single case study: how "masterpiece" can become "misterpieman" and "curiosity" "suretoy"

Formal lexical errors are relatively rare in the production of aphasic patients. In this study, we report the case of DW, who makes a high proportion of these errors. A few other cases have previously been reported, but DW shows a number of distinguishing characteristics. First, formal lexical errors are made in spelling and not in spoken speech. Second, they are associated with morphological errors and not with semantic errors. Third, they often combine lexical units in ways which are semantically and morphologically illegal. Finally, the majority of morphological errors involve the insertion, rather than the deletion, of suffixes. This pattern can be explained by hypothesizing that DW's errors arise because of confusions among a cohort of lexical neighbors activated top-down from a phonological input and bottom-up from shared letters. One possible cause of the confusions is lack of proper inhibition among lexical competitors.     

Disentangling the effects of word frequency and contextual diversity on serial recall performance

Research shows that contextual diversity (CD; the number of different contexts in which a word appears within a corpus) constitutes a better predictor of reading performance than word frequency (WF), that it mediates the access to lexical representations, and that controlling for contextual CD abolishes the effect of WF in lexical decision tasks. Despite the theoretical relevance of these findings for the study of serial memory, it is not known how CD might affect serial recall performance. We report the first independent manipulation of CD and WF in a serial recall task. Experiment 1 revealed better performance for low CD and for high WF words independently. Both effects affected omissions and item errors, but contrary to past research, word frequency also affected order errors. These results were confirmed in two more experiments comparing pure and alternating lists of low and high CD (Experiment 2) or WF (Experiment 3). The effect of CD was immune to this manipulation, while that of WF was abolished in alternating lists. Altogether the findings suggest a more difficult episodic retrieval of item information for words of high CD, and a role for both item and order information in the WF effect.     

A corpus-based analysis of verb continuation frequencies for syntactic processing

In the experimental study of how much verb guidance occurs in parsing syntactic ambiguity, the preference of an ambiguously subcategorized verb for its continuations is a relevant factor in the choice of the stimuli and in the interpretation of the results. Many methods have been used to measure this bias, ranging from experimenter's intuitions to sentence completion studies. In this paper, I assume that an acceptable definition of frequency is the count of the occurrences in a corpus, and I provide some cooccurrence counts calculated on a set of verbs and their syntactic siblings in a subset of the Penn Treebank (Marcus, Santorini, & Marcinkiewicz, 1993). Second, I perform some correlations with sentence completion and sentence production studies, which show that the two data collection methods are not strongly correlated with the corpus counts. Finally, I analyze the stimuli of some experiments which have shown evidence both in favor of the garden path theory of sentence processing and the lexical guidance theory. I argue that all the stimuli were balanced, and that a minimal attachment effect is not a consequence of the overall NP-bias of the stimuli. Moreover, evidence from the relation between processing times and verb completions with declarative and interrogative complementizers is used to argue that the parser is sensitive to the lexical content of the complementizers.     

Long-term priming of neighbours biases the word recognition process: evidence from a lexical decision task.

The role of orthographically similar words (i.e., neighbours) in the word recognition process has been studied extensively using short-term priming paradigms (e.g., Colombo, 1986). Here we demonstrate that long-term effects of neighbour priming can also be obtained. Experiment 1 showed that prior study of a neighbour (e.g., TANGO) increased later lexical decision performance for similar words (e.g., MANGO), but decreased performance for similar pseudowords (e.g., LANGO). Experiment 2 replicated this bias effect and showed that the increase in lexical decision performance due to neighbour priming is selectively due to words from a relatively sparse neighbourhood. Explanations of the bias effect in terms of lexical activation and episodic memory retrieval are discussed.     

Sublexical or lexical effects on serial recall of nonwords?

S. E. Gathercole, C. R. Frankish, S. J. Pickering, and S. Peaker (1999) reported 2 experiments in which they manipulated phonotactic properties of nonword stimuli and observed the effects on serial recall. Their results show superior recall for items consisting of more frequent phoneme pairs (biphone frequency). Biphone frequency was counted as the number of 3 phoneme words in which the phoneme pair occurs. In the first experiment of the current article, the authors made the same manipulation while controlling for the number of lexical neighbors and found no effect of biphone frequency. In the second experiment, the authors manipulated neighborhood size while controlling biphone frequency and found a significant effect of neighborhood size. The authors argued that serial recall of nonwords is influenced by lexical rather than sublexical knowledge.     

Are phonological influences on lexical (mis)selection the result of a monitoring bias?

A monitoring bias account is often used to explain speech error patterns that seem to be the result of an interactive language production system, like phonological influences on lexical selection errors. A biased monitor is suggested to detect and covertly correct certain errors more often than others. For instance, this account predicts that errors that are phonologically similar to intended words are harder to detect than those that are phonologically dissimilar. To test this, we tried to elicit phonological errors under the same conditions as those that show other kinds of lexical selection errors. In five experiments, we presented participants with high cloze probability sentence fragments followed by a picture that was semantically related, a homophone of a semantically related word, or phonologically related to the (implicit) last word of the sentence. All experiments elicited semantic completions or homophones of semantic completions, but none elicited phonological completions. This finding is hard to reconcile with a monitoring bias account and is better explained with an interactive production system. Additionally, this finding constrains the amount of bottom-up information flow in interactive models.     

Are phonological influences on lexical (mis)selection the result of a monitoring bias?

A monitoring bias account is often used to explain speech error patterns that seem to be the result of an interactive language production system, like phonological influences on lexical selection errors. A biased monitor is suggested to detect and covertly correct certain errors more often than others. For instance, this account predicts that errors that are phonologically similar to intended words are harder to detect than those that are phonologically dissimilar. To test this, we tried to elicit phonological errors under the same conditions as those that show other kinds of lexical selection errors. In five experiments, we presented participants with high cloze probability sentence fragments followed by a picture that was semantically related, a homophone of a semantically related word, or phonologically related to the (implicit) last word of the sentence. All experiments elicited semantic completions or homophones of semantic completions, but none elicited phonological completions. This finding is hard to reconcile with a monitoring bias account and is better explained with an interactive production system. Additionally, this finding constrains the amount of bottom-up information flow in interactive models.     

Modeling lexical decision: the form of frequency and diversity effects

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