Commentary on “Interaction in Spoken Word Recognition Models”

Norris, Dennis; McQueen, James M.; Cutler, Anne

doi:10.3389/fpsyg.2018.01568

Cited by 10 publications

(9 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, an individual will report hearing a speech sound that has been replaced by noise (e.g., hearing the phoneme /s/ in the frame legi_lature , where the critical phoneme has been replaced by a cough; Warren, 1970), and an individual's estimate of an object's size can be influenced by the width between a person's hands (Stefanucci & Geuss, 2009). While such context effects are ubiquitous across a range of domains in cognitive psychology, an ongoing debate––whether in the domain of language (e.g., Magnuson, Mirman, Luthra, Strauss, & Harris, 2018; Norris, McQueen, & Cutler, 2018) or the domain of vision (e.g., Firestone & Scholl, 2014, 2016; Gilbert & Li, 2013; Lupyan, Abdel Rahman, Boroditsky, & Clark, 2020; Schnall, 2017a, 2017b)––centers on how contextual information is integrated with sensory signals. In particular, do contextual effects on sensory processing reflect influences on perception itself, or does context only affect an individual's postperceptual decisions?…”

Section: Introductionmentioning

confidence: 99%

Robust Lexically Mediated Compensation for Coarticulation: Christmash Time Is Here Again

et al. 2021

View full text Add to dashboard Cite

A long‐standing question in cognitive science is how high‐level knowledge is integrated with sensory input. For example, listeners can leverage lexical knowledge to interpret an ambiguous speech sound, but do such effects reflect direct top‐down influences on perception or merely postperceptual biases? A critical test case in the domain of spoken word recognition is lexically mediated compensation for coarticulation (LCfC). Previous LCfC studies have shown that a lexically restored context phoneme (e.g., /s/ in Christma#) can alter the perceived place of articulation of a subsequent target phoneme (e.g., the initial phoneme of a stimulus from a tapes‐capes continuum), consistent with the influence of an unambiguous context phoneme in the same position. Because this phoneme‐to‐phoneme compensation for coarticulation is considered sublexical, scientists agree that evidence for LCfC would constitute strong support for top–down interaction. However, results from previous LCfC studies have been inconsistent, and positive effects have often been small. Here, we conducted extensive piloting of stimuli prior to testing for LCfC. Specifically, we ensured that context items elicited robust phoneme restoration (e.g., that the final phoneme of Christma# was reliably identified as /s/) and that unambiguous context‐final segments (e.g., a clear /s/ at the end of Christmas) drove reliable compensation for coarticulation for a subsequent target phoneme. We observed robust LCfC in a well‐powered, preregistered experiment with these pretested items (N = 40) as well as in a direct replication study (N = 40). These results provide strong evidence in favor of computational models of spoken word recognition that include top–down feedback.

show abstract

Section: Introductionmentioning

confidence: 99%

Robust Lexically Mediated Compensation for Coarticulation: Christmash Time Is Here Again

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Indeed, Norris et al (2018) make precisely this claim, and say the only reason feedback helps in TRACE is "because its initial behavior is suboptimal". Although we disagree about "suboptimality"…”

Section: Discussionmentioning

confidence: 93%

“…Suboptimal behavior is not necessarily incorrect behavior; the interesting question is to what degree a model corresponds with humans at behavioral, algorithmic, and neural levels (Magnuson et al, 2018), and optimality is an interesting baseline to consider. Norris et al (2018) argue that "...the best that any speech recognition system can do is compute the match between input features and lexical representations and select the best-matching word (more specifically, pick the word with maximum likelihood) ... Shortlist B...by virtue of implementing Bayesian inference, performs optimally; its use of Bayes' rule guarantees that the best-matching word must be recognized (p. 1)". Some details here may help us better understand and possibly refine the debate between autonomous (without feedback) and interactive (with feedback) accounts.…”

Section: Discussionmentioning

confidence: 99%

“…Recently, our team laid out a comprehensive case, via argument and demonstrations with model simulations, for how feedback can promote faster processing and make recognition robust in noise (Magnuson et al, 2018). In response, Norris, McQueen, and Cutler (2018) claimed that demonstrating a role for feedback in TRACE would be inconsequential (because it might not help in other models) and that arguments on the basis of parsimony were problematic without formal model comparisons. Feedback, they argued, should only be added to models of spoken word recognition if it is necessary to do so to account for human performance, and they emphasized that there was not conclusive evidence for feedback in behavioral studies with human subjects.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Contra assertions, feedback improves word recognition

Magnuson¹,

Grubb²,

Crinnion³

et al. 2021

Preprint

View full text Add to dashboard Cite

Norris and Cutler (in press) revisit their arguments that (lexical-to-sublexical) feedback cannot improve word recognition performance, based on the assumption that feedback must boost signal and noise equally. They also argue that demonstrations that feedback improves performance (Magnuson, Mirman, Luthra, Strauss, & Harris, 2018) in the TRACE model of spoken word recognition (McClelland & Elman, 1986) were artifacts of converting activations to response probabilities. We first evaluate their claim that feedback in an interactive activation model must boost noise and signal equally. This is not true in a fully interactive activation model such as TRACE, where the feedback signal does not simply mirror the feedforward signal; it is instead shaped by joint probabilities over lexical patterns, and the dynamics of lateral inhibition. Thus, even under high levels of noise, lexical feedback will selectively boost signal more than noise. We demonstrate that feedback promotes faster word recognition and preserves accuracy under noise whether one uses raw activations or response probabilities. We then document that lexical feedback selectively boosts signal (i.e., lexically-coherent series of phonemes) more than noise by tracking sublexical (phoneme) activations under noise with and without feedback. Thus, feedback in a model like TRACE does improve word recognition, exactly by selective reinforcement of lexically-coherent signal. We conclude that whether lexical feedback is integral to human speech processing is an empirical question, and briefly review a growing body of work at behavioral and neural levels that is consistent with feedback and inconsistent with autonomous (non-feedback) architectures.

show abstract

“…The details of the cognitive processes underlying human speech processing are still contentious. A long-standing debate revolves around the importance and timing of top-down versus bottom-up influence for word recognition during speech comprehension [2,3]. Certain autonomous models (e.g.…”

Section: Introductionmentioning

confidence: 99%

Quantifying Expectation Modulation in Human Speech Processing

Bentum¹,

Bosch²,

Bosch

et al. 2019

Interspeech 2019

View full text Add to dashboard Cite

The mismatch between top-down predicted and bottom-up perceptual input is an important mechanism of perception according to the predictive coding framework (Friston, [1]). In this paper we develop and validate a new information-theoretic measure that quantifies the mismatch between expected and observed auditory input during speech processing. We argue that such a mismatch measure is useful for the study of speech processing. To compute the mismatch measure, we use naturalistic speech materials containing approximately 50,000 word tokens. For each word token we first estimate the prior word probability distribution with the aid of statistical language modelling, and next use automatic speech recognition to update this word probability distribution based on the unfolding speech signal. We validate the mismatch measure with multiple analyses, and show that the auditory-based update improves the probability of the correct word and lowers the uncertainty of the word probability distribution. Based on these results, we argue that it is possible to explicitly estimate the mismatch between predicted and perceived speech input with the cross entropy between word expectations computed before and after an auditory update.

show abstract

Commentary on “Interaction in Spoken Word Recognition Models”

Cited by 10 publications

References 13 publications

Robust Lexically Mediated Compensation for Coarticulation: Christmash Time Is Here Again

Robust Lexically Mediated Compensation for Coarticulation: Christmash Time Is Here Again

Contra assertions, feedback improves word recognition

Quantifying Expectation Modulation in Human Speech Processing

Contact Info

Product

Resources

About