Shift In Category Boundary Research Articles

Category boundaries modulate memory in a place-cell-like manner

Concepts describe how instances of the same kind are related, enabling the categorization and interpretation of new information.1,2 How concepts are represented is a longstanding question. Category boundaries have been considered defining features of concept representations, which can guide categorical inference,3 with fMRI evidence showing category-boundary signals in the hippocampus.4,5 The underlying neural mechanism remains unclear. The hippocampal-entorhinal system, known for its spatially tuned neurons that form cognitive maps of space,6,7 may support conceptual knowledge formation, with place cells encoding locations in conceptual space.4,8,9,10,11 Physical boundaries anchor spatial representations and boundary shifts affect place and grid fields,12,13,14,15,16 as well as human spatial memory,17,18,19 along manipulated dimensions. These place cell responses are likely driven by boundary vector cells, which respond to boundaries at specific allocentric distances and directions,20,21,22,23 the neural correlates of which have been identified in the subiculum and entorhinal cortex20,24,25. We hypothesize similar patterns of memory adaptations in response to shifting category boundaries. Our findings show that after category boundary shifts, participants' memory for category exemplars distorts along the changed dimension, mirroring place field deformations. We demonstrate that the boundary vector cell model of place cell firing best accounts for these distortions compared with alternative geometric explanations. Our study highlights a role of category boundaries in human cognition and establishes a new complementary link between hippocampal coding properties with respect to boundaries and human concept representation, bridging spatial and conceptual domains.

Contextual effects on the distinction of Japanese length contrasts in reverberation

Temporal structure in a word is crucial for understanding speech. In the distinction between Japanese singleton and geminate stops, listeners tended to perceive a consonant as geminate when the preceding vowel was longer [Ofuka et al., J. Phonetics Soc. Japan 9(2), 59–65 (2005)]. At the same time, listeners may not be able to use the contextual information in reverberation since the boundaries between each speech sound may become blurred by reverberation elongating and overlapping the sounds. The current study investigated whether the category boundary between singleton and geminate might change depending on duration of the preceding vowel in reverberation. Participants made a choice between singleton and geminate consonants embedded in an /a a/ context. The duration of the preceding vowel was changed. The experiments were conducted in a reverberant (RT = 2.6 s) and non-reverberant (RT = 0.1 s) condition. The results showed a significant shift in category boundary which indicated that listeners gave more “geminate” responses when the preceding vowel was longer, especially in the reverberant condition. The preceding vowel lengthened by reverberation caused listeners to choose “geminate.” Listeners might be more sensitive to the length of the adjacent phonemes for assessing the degree of reverberation.Temporal structure in a word is crucial for understanding speech. In the distinction between Japanese singleton and geminate stops, listeners tended to perceive a consonant as geminate when the preceding vowel was longer [Ofuka et al., J. Phonetics Soc. Japan 9(2), 59–65 (2005)]. At the same time, listeners may not be able to use the contextual information in reverberation since the boundaries between each speech sound may become blurred by reverberation elongating and overlapping the sounds. The current study investigated whether the category boundary between singleton and geminate might change depending on duration of the preceding vowel in reverberation. Participants made a choice between singleton and geminate consonants embedded in an /a a/ context. The duration of the preceding vowel was changed. The experiments were conducted in a reverberant (RT = 2.6 s) and non-reverberant (RT = 0.1 s) condition. The results showed a significant shift in category boundary which indicated that listeners gave more ...

Consonant categorization exhibits a graded influence of surrounding spectral context.

When spectral properties differ across successive sounds, this difference is perceptually magnified, resulting in spectral contrast effects (SCEs). Recently, Stilp, Anderson, and Winn [(2015) J. Acoust. Soc. Am. 137(6), 3466-3476] revealed that SCEs are graded: more prominent spectral peaks in preceding sounds produced larger SCEs (i.e., category boundary shifts) in categorization of subsequent vowels. Here, a similar relationship between spectral context and SCEs was replicated in categorization of voiced stop consonants. By generalizing this relationship across consonants and vowels, different spectral cues, and different frequency regions, acute and graded sensitivity to spectral context appears to be pervasive in speech perception.

More than a boundary shift: Perceptual adaptation to foreign-accented speech reshapes the internal structure of phonetic categories.

The literature on perceptual learning for speech shows that listeners use lexical information to disambiguate phonetically ambiguous speech sounds and that they maintain this new mapping for later recognition of ambiguous sounds for a given talker. Evidence for this kind of perceptual reorganization has focused on phonetic category boundary shifts. Here, we asked whether listeners adjust both category boundaries and internal category structure in rapid adaptation to foreign accents. We investigated the perceptual learning of Mandarin-accented productions of word-final voiced stops in English. After exposure to a Mandarin speaker's productions, native-English listeners' adaptation to the talker was tested in 3 ways: a cross-modal priming task to assess spoken word recognition (Experiment 1), a category identification task to assess shifts in the phonetic boundary (Experiment 2), and a goodness rating task to assess internal category structure (Experiment 3). Following exposure, both category boundary and internal category structure were adjusted; moreover, these prelexical changes facilitated subsequent word recognition. Together, the results demonstrate that listeners' sensitivity to acoustic-phonetic detail in the accented input promoted a dynamic, comprehensive reorganization of their perceptual response as a consequence of exposure to the accented input. We suggest that an examination of internal category structure is important for a complete account of the mechanisms of perceptual learning. (PsycINFO Database Record

Accounting for rate-dependent category boundary shifts in speech perception

The perception of temporal contrasts in speech is known to be influenced by the speech rate in the surrounding context. This rate-dependent perception is suggested to involve general auditory processes because it is also elicited by nonspeech contexts, such as pure tone sequences. Two general auditory mechanisms have been proposed to underlie rate-dependent perception: durational contrast and neural entrainment. This study compares the predictions of these two accounts of rate-dependent speech perception by means of four experiments, in which participants heard tone sequences followed by Dutch target words ambiguous between /ɑs/ “ash” and /a:s/ “bait”. Tone sequences varied in the duration of tones (short vs. long) and in the presentation rate of the tones (fast vs. slow). Results show that the duration of preceding tones did not influence target perception in any of the experiments, thus challenging durational contrast as explanatory mechanism behind rate-dependent perception. Instead, the presentation rate consistently elicited a category boundary shift, with faster presentation rates inducing more /a:s/ responses, but only if the tone sequence was isochronous. Therefore, this study proposes an alternative, neurobiologically plausible account of rate-dependent perception involving neural entrainment of endogenous oscillations to the rate of a rhythmic stimulus.

Differences in perceptual learning transfer as a function of training task.

A growing body of research--including results from behavioral psychology, human structural and functional imaging, single-cell recordings in nonhuman primates, and computational modeling--suggests that perceptual learning effects are best understood as a change in the ability of higher-level integration or association areas to read out sensory information in the service of particular decisions. Work in this vein has argued that, depending on the training experience, the "rules" for this read-out can either be applicable to new contexts (thus engendering learning generalization) or can apply only to the exact training context (thus resulting in learning specificity). Here we contrast learning tasks designed to promote either stimulus-specific or stimulus-general rules. Specifically, we compare learning transfer across visual orientation following training on three different tasks: an orientation categorization task (which permits an orientation-specific learning solution), an orientation estimation task (which requires an orientation-general learning solution), and an orientation categorization task in which the relevant category boundary shifts on every trial (which lies somewhere between the two tasks above). While the simple orientation-categorization training task resulted in orientation-specific learning, the estimation and moving categorization tasks resulted in significant orientation learning generalization. The general framework tested here--that task specificity or generality can be predicted via an examination of the optimal learning solution--may be useful in building future training paradigms with certain desired outcomes.

A following sibilant increases the ambiguity of a sibilant continuum

We examined the effect of three following contexts: /s/, /∫/, and vowel, on the categorization of a /s/-/∫/ continuum. Unlike previous findings of a shift in category boundary due to context [Mann and Repp (1980)], we found that in the context of a following sibilant, listeners found the target sibilant to be more ambiguous (shallower categorization slopes and responses closer to chance) than when followed by a vowel (p < 0.001). There was also a tendency for the /∫/ context (which affects pronunciation) to create more ambiguity than the /s/ context (which does not) (p = 0.057). In a second experiment, half the participants heard the following context as part of the same syntactic phrase as the target (e.g., “Whenever they fra? Shelly gets upset”) and half heard it as part of a different phrase (e.g., “Whenever they fra? Shelly, John gets upset”). Pronunciation usually is more affected when target and context are in the same phrase (Holst and Nolan (1995). Although our stimuli were acoustically identical, listeners tended to perceive target sibilants as more ambiguous when the following /∫/ was part of the same phrase (p = 0.059), suggesting a role for top-down knowledge in interpreting segmental information.

The impact of tone systems on the categorical perception of lexical tones: An event-related potentials study

This study investigates the categorical perception (CP) of pitch contours (level and rising) by native listeners of two tone languages, Mandarin and Cantonese, for both speech and nonspeech. Language background was found to modulate participants’ behavioural and electrophysiological responses to stimuli presented in an active oddball paradigm, comprising a standard and two equally spaced deviants (within- and across-category). The stimuli were divided into two sets according to the results of a two-alternative forced-choice identification test: a rising set, using a standard that listeners identified as high rising tone, and a level set, using a standard that listeners identified as high level tone. For the rising set, both groups of listeners exhibited CP in terms of their behavioural response. However, only Cantonese listeners exhibited a significant CP effect in terms of P300 amplitude. For the level set, the behavioural data revealed a shift in category boundary due, in part, to the range–frequency effect. According to the d′ scores, the CP effect elicited from Mandarin listeners was greater for nonspeech stimuli than for speech, suggesting the presence of apsychophysical boundary. There was no such behavioural contrast for Cantonese listeners. However, Cantonese listeners exhibited a significant CP effect in P300 amplitude that was influenced by the range–frequency effect, as well as a possible secondary phonological boundary. P300 amplitude is believed to index the ease of discrimination of speech stimuli by phonological information. We conclude that Cantonese listeners engaged phonological processing in order to discriminate speech stimuli more efficiently than Mandarin listeners. These findings may be due to the different tonal inventories of Mandarin and Cantonese, with Cantonese listeners required to make finer distinctions in perception of pitch height and slope than Mandarin listeners in order to discriminate the denser tone system of Cantonese.

Flexible Categorization of Relative Stimulus Strength by the Optic Tectum

Categorization is the process by which the brain segregates continuously variable stimuli into discrete groups. We report that patterns of neural population activity in the owl optic tectum (OT) categorize stimuli based on their relative strengths into "strongest" versus "other." The category boundary shifts adaptively to track changes in the absolute strength of the strongest stimulus. This population-wide categorization is mediated by the responses of a small subset of neurons. Our data constitute the first direct demonstration of explicit categorization of stimuli by a neural network based on relative stimulus strength or salience. The finding of categorization by the population code relaxes constraints on the properties of downstream decoders that might read out the location of the strongest stimulus. These results indicate that the ensemble neural code in the OT could mediate bottom-up stimulus selection for gaze and attention, a form of stimulus categorization in which the category boundary often shifts within hundreds of milliseconds.

The Neural Bases of the Lexical Effect: An fMRI Investigation

The lexical effect is a phenomenon whereby lexical information influences the perception of the phonetic category boundary for stimuli from word-nonword continua. At issue is whether this effect is due to "top-down" influence of upper levels of processing on perceptual processing, or instead is due to decision-stage processes. In this study, brain activity was monitored using functional magnetic resonance imaging as subjects performed a phonetic categorization task on items taken from 2 continua in which one end of the continuum was a real word and the other was not (gift-kift and giss-kiss). If the lexical effect has a perceptual basis, modulation of activation should be seen as a function of the lexical effect in areas such as the superior temporal gyri (STG) which have previously been implicated in perceptual processing. In contrast, if the effect is purely due to decision-related factors, such modulation would be expected only in areas which have been linked to executive processes, such as frontal and midline structures. Modulation of activation as a function of the lexically biased shift in phonetic category boundary was observed in the STG bilaterally as well as in frontal and midline structures. This activation pattern suggests that the lexical effect has at minimum a perceptual component, in addition to an executive decision-related component. These results challenge the view that lexical effects on phonetic boundary placement are due solely to postperceptual, decision-stage processes, and support those models of language processing which allow for higher-level lexical information to directly influence the perception of incoming speech.

Effects of frequency shifts and visual gender information on vowel category judgments

Visual morphing techniques were used together with a high‐quality vocoder to study the audiovisual contribution of talker gender to the identification of frequency‐shifted vowels. A nine‐step continuum ranging from ‘‘bit’’ to ‘‘bet’’ was constructed from natural recorded syllables spoken by an adult female talker. Upward and downward frequency shifts in spectral envelope (scale factors of 0.85 and 1.0) were applied in combination with shifts in fundamental frequency, F0 (scale factors of 0.5 and 1.0). Downward frequency shifts generally resulted in malelike voices whereas upward shifts were perceived as femalelike. Two separate nine‐step visual continua from ‘‘bit’’ to ‘‘bet’’ were also constructed, one from a male face and the other a female face, each producing the end‐point words. Each step along the two visual continua was paired with the corresponding step on the acoustic continuum, creating natural audiovisual utterances. Category boundary shifts were found for both acoustic cues (F0 and formant fre...

Language-specific patterns of vowel-to-vowel coarticulation: acoustic structures and their perceptual correlates

Three experiments tested the hypothesis that V-to-V coarticulatory organization differs in Shona and English, and that Shona- and English-speaking listeners' sensitivity to V-to-V coarticulatory effects is correspondingly language-specific. An acoustic study of Shona and English CV′CVCV trisyllables (Experiment 1) showed that the two languages differ in the carryover vs. anticipatory influences of stressed and unstressed vowels on each other. In 4IAX discrimination tests in which both Shona and English coarticulatory effects were spliced into different coarticulatory contexts (Experiment 2), Shona and English listeners perceptually compensated more (i.e., attributed more of a vowel's acoustic properties to its coarticulatory context in targeted test trials) for effects that were consistent with their linguistic experience. Similarly, when these listeners identified synthetic target vowels embedded into different vowel contexts (Experiment 3), Shona listeners compensated more (i.e., showed larger category boundary shifts) for the vowel contexts that triggered larger acoustic influences in the production study. English listeners' boundary shifts were more complicated but, when these data were combined with those from a follow-up identification study, they showed the perceptual shifts expected on the basis of the English coarticulatory findings. Overall, the relation between the production and perception data suggests that listeners are attuned to native-language coarticulatory patterns.

A sequential contrast effect in odor perception

An odor of ambiguous quality was shown to shift in rated odor character after presentation of more prototypical odors, always in a direction contrasting with the previous context. The terpene aroma compound, dihydromyrcenol, is perceived as partially woody (pine-like) and partially citrus (lime-like) in odor character. Citrus ratings of this odor increased following exposure to woody odors. Conversely, woody ratings of dihydromyrcenol increased following exposure to citrus odors. Possible explanations for this sequential contrast effect include shifts in cognitive category boundaries, response frequency biases, and simple sensory adaptation.

Effects of listening to repeated syllables: category boundary shifts versus verbal transformations

Listening to recordings of clearly pronounced repeated syllables can produce quite different perceptual effects depending upon whether observations are perstimulatory (during stimulation) or poststimulatory (after stimulation ceases). Poststimulatory changes involve migration of category boundaries separating speech sounds—for example a shift in the /t/–/d/ voice onset time (VOT) boundary toward that of a previously repeated consonant–vowel syllable such as “to” or “do”. Studies of the perstimulatory changes called “verbal transformations” usually employ phonetically complex stimuli, and consist of abrupt illusory changes in what is heard, frequently accompanied by considerable phonetic distortion. The present study examined verbal transformations reported while listening to alternating statements of “to” and “do” (“to, do, to, do …”)for which the adaptational VOT boundary changes produced by one syllable opposed that of the other. The perstimulatory reports were inconsistent with adaptational category boundary shifts. It is suggested that although poststimulatory changes reflect a criterion shift rule applicable to perceptual judgments in general, perstimulatory verbal transformations involve perceptual cleavage into organizational units, some of which mutate before recombining to yield a different linguistic form. Verbal transformations provide information concerning linguistic processing which differs from that furnished by poststimulatory changes.

Effects of listening to repeated syllables: Perstimulatory verbal transformations versus poststimulatory shifts in category boundaries

Listening to repeated syllables produces two quite different effects depending upon whether observations are made following or during repetition. Poststimulatory changes in category boundaries (for example, shifts in the voice‐onset time boundary separation /tu/ from /du/ after hearing repetitions of either /tu/ or /du/) have been the subject of lively controversy concerning whether or not feature‐detector adaptation (FDA) is responsible for such changes. Curiously, studies dealing with poststimulatory effects have not considered perstimulatory effects developing during stimulation. It is known that illusory changes (verbal transformations) occur while listening to repeated syllables, and the present study examined these verbal transformations for evidence of FDA. The repeated stimulus …/tu/, /du/, /tu/, /du/…, was used, since little or no net boundary change should be produced by FDA. The verbal transformation reported were inconsistent with adaptation theories. It is suggested that poststimulatory effects follow a “criterion shift rule” applying to perception in general, while verbal transformations tap special linquistic mechanisms employed normally to correct errors and enhance intelligibility of speech. [Work supported by NIH and NSF.]

Identification of synthesized [r-w] continua for adult and child speakers.

A continuing problem in speech research is the comparison of phonetically similar but acoustically distinct sounds such as the comparison between adult speech and child speech. This problem is especially important in the remediation of speech disorders. In this study, we focus on this problem as manifested in the disordered production of [r] sounds. Stimuli which varied in F2 and F3 onset frequencies were synthesized to span continua from [r] to [w] for adult, child, and scaled adult (formants 1.5 times adult values) speakers. These stimuli were presented to 18 naive adults for identification as "R"or "W" at three listening sessions to test for stability of category boundaries. Significant shifts in category boundaries were found for all three continua between sessions one and two but not between sessions two and three. Category boundaries in terms of formant frequencies showed significant differences between the adult and childlike continua. Rescaling these values on a psychophysical dimension and adjusting for formant spread brought the category boundaries into close alignment. In free choice testing, synthesized [r] and [w] stimuli were identified by most subjects at least as well as natural [r] and [w] stimuli but synthesized child [r] stimuli were identified by most subject considerably better than natural child [r] stimuli.

Influence of preceding [s] or [∫] on stop consonant perception

The category boundaries on synthetic [da]-[ga] and [du]-[gu] continua are shifted substantially when the syllables are preceded by one of the two synthetic fricatives, [s] or [∫]: More velar stops are heard following [s] than following [∫]. We have investigated how this contextual dependency changes as (1) the silent closure period following the fricative is extended, and as (2) the fricative is preceded by a vowel to create a syllable boundary between fricative and stop. Preliminary data indicate that, while neither manipulation completely removes the contextual effect, both cause a reduction in its size. We suppose that the context dependency reflects the dynamics of speech production. Fricative-induced shifts in stop consonant category boundaries presumably correspond to coarticulatory shifts in the precise place of contact between tongue and palate. We expect to obtain measurements on natural speech that bear on this hypothesis. [Work supported by NICHD and BRSG.]