Context-dependent Encoding Research Articles

Testing models of context-dependent outcome encoding in reinforcement learning

Previous studies of reinforcement learning (RL) have established that choice outcomes are encoded in a context-dependent fashion. Several computational models have been proposed to explain context-dependent encoding, including reference point centering and range adaptation models. The former assumes that outcomes are centered around a running estimate of the average reward in each choice context, while the latter assumes that outcomes are compared to the minimum reward and then scaled by an estimate of the range of outcomes in each choice context. However, there are other computational mechanisms that can explain context dependence in RL. In the present study, a frequency encoding model is introduced that assumes outcomes are evaluated based on their proportional rank within a sample of recently experienced outcomes from the local context. A range-frequency model is also considered that combines the range adaptation and frequency encoding mechanisms. We conducted two fully incentivized behavioral experiments using choice tasks for which the candidate models make divergent predictions. The results were most consistent with models that incorporate frequency or rank-based encoding. The findings from these experiments deepen our understanding of the underlying computational processes mediating context-dependent outcome encoding in human RL.

More evidence against the Spinozan model: Cognitive load diminishes memory for "true" feedback.

We tested two competing models on the memory representation of truth-value information: the Spinozan model and the Cartesian model. Both models assume that truth-value information is represented with memory "tags," but the models differ in their coding scheme. According to the Cartesian model, true information is stored with a "true" tag, and false information is stored with a "false" tag. In contrast, the Spinozan model proposes that only false information receives "false" tags. All other (i.e., untagged) information is considered as true by default. Hence, in case of cognitive load during feedback encoding, the latter model predicts a load effect on memory for "false" feedback, but not on memory for "true" feedback. To test this prediction, participants studied trivia statements (Experiment 1) or nonsense statements that allegedly represented foreign-language translations (Experiment 2). After each statement, participants received feedback on the (alleged) truth value of the statement. Importantly, half of the participants experienced cognitive load during feedback processing. For the trivia statements of Experiment 1, we observed a load effect on memory for both "false" and "true" feedback. In contrast, for the nonsense statements of Experiment 2, we found a load effect on memory for "true" feedback only. Both findings clearly contradict the Spinozan model. However, our results are also only partially in line with the predictions of the Cartesian model. For this reason, we suggest a more flexible model that allows for an optional and context-dependent encoding of "true" tags and "false" tags.

Interactive effects of linguistic abstraction and stimulus statistics in the online modulation of neural speech encoding.

Speech processing is highly modulated by context. Prior studies examining frequency-following responses (FFRs), an electrophysiological 'neurophonic' potential that faithfully reflects phase-locked activity from neural ensembles within the auditory network, have demonstrated that stimulus context modulates the integrity of speech encoding. The extent to which context-dependent encoding reflects general auditory properties or interactivities between statistical and higher-level linguistic processes remains unexplored. Our study examined whether speech encoding, as reflected by FFRs, is modulated by abstract phonological relationships between a stimulus and surrounding contexts. FFRs were elicited to a Mandarin rising-tone syllable (/ji-TR/, 'second') randomly presented with other syllables in three contexts from 17 native listeners. In a contrastive context, /ji-TR/ occurred with meaning-contrastive high-level-tone syllables (/ji-H/, 'one'). In an allotone context, TR occurred with dipping-tone syllables /ji-D/, a non-meaning-contrastive variant of /ji-TR/. In a repetitive context, the same /ji-TR/ occurred with other speech tokens of /ji-TR/. Consistent with prior work, neural tracking of /ji-TR/ pitch contour was more faithful in the repetitive condition wherein /ji-TR/ occurred more predictably (p = 1) than in the contrastive condition (p = 0.34). Crucially, in the allotone context, neural tracking of /ji-TR/ was more accurate relative to the contrastive context, despite both having an identical transitional probability (p = 0.34). Mechanistically, the non-meaning-contrastive relationship may have augmented the probability to /ji-TR/ occurrence in the allotone context. Results indicate online interactions between bottom-up and top-down mechanisms, which facilitate speech perception. Such interactivities may predictively fine-tune incoming speech encoding using linguistic and statistical information from prior context.

Encoding of Articulatory Kinematic Trajectories in Human Speech Sensorimotor Cortex

When speaking, we dynamically coordinate movements of our jaw, tongue, lips, and larynx. To investigate the neural mechanisms underlying articulation, we used direct cortical recordings from human sensorimotor cortex while participants spoke natural sentences that included sounds spanning the entire English phonetic inventory. We used deep neural networks to infer speakers' articulator movements from produced speech acoustics. Individual electrodes encoded a diversity of articulatory kinematic trajectories (AKTs), each revealing coordinated articulator movements toward specific vocal tract shapes. AKTs captured a wide range of movement types, yet they could be differentiated by the place of vocal tract constriction. Additionally, AKTs manifested out-and-back trajectories with harmonic oscillator dynamics. While AKTs were functionally stereotyped across different sentences, context-dependent encoding of preceding and following movements during production of the same phoneme demonstrated the cortical representation of coarticulation. Articulatory movements encoded in sensorimotor cortex give rise to the complex kinematics underlying continuous speech production. VIDEO ABSTRACT.

Acoustic basis of context dependent brainstem encoding of speech

The newfound context dependent brainstem encoding of speech is evidence of online regularity detection and modulation of the sub-cortical responses. We studied the influence of spectral structure of the contextual stimulus on context dependent encoding of speech at the brainstem, in an attempt to understand the acoustic basis for this effect. Fourteen normal hearing adults participated in a randomized true experimental design in whom brainstem responses were recorded. Brainstem responses for a high pass filtered /da/ in the context of syllables, that either had same or different spectral structure were compared with each other. The findings suggest that spectral structure is one of the parameters which cue the context dependent sub-cortical encoding of speech. Interestingly, the results also revealed that, brainstem can encode pitch even with negligible acoustic information below the second formant frequency.

Context-dependent encoding in the auditory brainstem subserves enhanced speech-in-noise perception in musicians

Musical training strengthens speech perception in the presence of background noise. Given that the ability to make use of speech sound regularities, such as pitch, underlies perceptual acuity in challenging listening environments, we asked whether musicians’ enhanced speech-in-noise perception is facilitated by increased neural sensitivity to acoustic regularities. To this aim we examined subcortical encoding of the same speech syllable presented in predictable and variable conditions and speech-in-noise perception in 31 musicians and nonmusicians. We anticipated that musicians would demonstrate greater neural enhancement of speech presented in the predictable compared to the variable condition than nonmusicians. Accordingly, musicians demonstrated more robust neural encoding of the fundamental frequency (i.e., pitch) of speech presented in the predictable relative to the variable condition than nonmusicians. The degree of neural enhancement observed to predictable speech correlated with subjects’ musical practice histories as well as with their speech-in-noise perceptual abilities. Taken together, our findings suggest that subcortical sensitivity to speech regularities is shaped by musical training and may contribute to musicians’ enhanced speech-in-noise perception.

Biological Markers of Reading and Speech-in-Noise Perception in the Auditory System

Biological Markers of Reading and Speech-in-Noise Perception in the Auditory System

Context-Dependent Encoding in the Human Auditory Brainstem Relates to Hearing Speech in Noise: Implications for Developmental Dyslexia

We examined context-dependent encoding of speech in children with and without developmental dyslexia by measuring auditory brainstem responses to a speech syllable presented in a repetitive or variable context. Typically developing children showed enhanced brainstem representation of features related to voice pitch in the repetitive context, relative to the variable context. In contrast, children with developmental dyslexia exhibited impairment in their ability to modify representation in predictable contexts. From a functional perspective, we found that the extent of context-dependent encoding in the auditory brainstem correlated positively with behavioral indices of speech perception in noise. The ability to sharpen representation of repeating elements is crucial to speech perception in noise, since it allows superior "tagging" of voice pitch, an important cue for segregating sound streams in background noise. The disruption of this mechanism contributes to a critical deficit in noise-exclusion, a hallmark symptom in developmental dyslexia.

Neurophysiological evidence for context-dependent encoding of sensory input in human auditory cortex

Attention biases the way in which sound information is stored in auditory memory. Little is known, however, about the contribution of stimulus-driven processes in forming and storing coherent sound events. An electrophysiological index of cortical auditory change detection (mismatch negativity [MMN]) was used to assess whether sensory memory representations could be biased toward one organization over another (one or two auditory streams) without attentional control. Results revealed that sound representations held in sensory memory biased the organization of subsequent auditory input. The results demonstrate that context-dependent sound representations modulate stimulus-dependent neural encoding at early stages of auditory cortical processing.