The use of sequential probabilities in the segmentation of speech

The present investigation addresses the possible utility of sequential probabilities in the segmentation of spoken language. In a series of five word- spotting and two control lexical decision experiments, high- versus low-probability consonant-vowel (Experiments 1, 2, 5, and 7) and vowel-consonant (Experiments 1, 3, 4, and 6) strings were presented either in the nonsense contexts of target words (Experiments 1–3) or within the target words themselves (Experiments 4–7). The results suggest that listeners, at least for sequences in the onset position, indeed use sequential probabilities as cues for segmentation. The probability of a sound sequence influenced segmentation more when the sequence occurred within the target words (Experiments 4–7 vs. Experiments 1–3). Furthermore, the effects were reliable only when the sequences occurred in the onset position (Experiments 1, 2, 5, and 7 vs. Experiments 1, 3, 4, and 6).     

The metamorphosis of the statistical segmentation output: Lexicalization during artificial language learning

This study combined artificial language learning (ALL) with conventional experimental techniques to test whether statistical speech segmentation outputs are integrated into adult listeners’ mental lexicon. Lexicalization was assessed through inhibitory effects of novel neighbors (created by the parsing process) on auditory lexical decisions to real words. Both immediately after familiarization and post-one week, ALL outputs were lexicalized only when the cues available during familiarization (transitional probabilities and wordlikeness) suggested the same parsing (Experiments 1 and 3). No lexicalization effect occurred with incongruent cues (Experiments 2 and 4). Yet, ALL differed from chance, suggesting a dissociation between item knowledge and lexicalization. Similarly contrasted results were found when frequency of occurrence of the stimuli was equated during familiarization (Experiments 3 and 4). Our findings thus indicate that ALL outputs may be lexicalized as far as the segmentation cues are congruent, and that this process cannot be accounted for by raw frequency.     

Lexical viability constraints on speech segmentation by infants

The Possible Word Constraint limits the number of lexical candidates considered in speech recognition by stipulating that input should be parsed into a string of lexically viable chunks. For instance, an isolated single consonant is not a feasible word candidate. Any segmentation containing such a chunk is disfavored. Five experiments using the head-turn preference procedure investigated whether, like adults, 12-month-olds observe this constraint in word recognition. In Experiments 1 and 2, infants were familiarized with target words (e.g., rush), then tested on lists of nonsense items containing these words in "possible" (e.g., "niprush" [nip+rush]) or "impossible" positions (e.g., "prush" [p+rush]). The infants listened significantly longer to targets in "possible" versus "impossible" contexts when targets occurred at the end of nonsense items (rush in "prush"), but not when they occurred at the beginning (tan in "tance"). In Experiments 3 and 4, 12-month-olds were similarly familiarized with target words, but test items were real words in sentential contexts (win in "wind" versus "window"). The infants listened significantly longer to words in the "possible" condition regardless of target location. Experiment 5 with targets at the beginning of isolated real words (e.g., win in "wind") replicated Experiment 2 in showing no evidence of viability effects in beginning position. Taken together, the findings suggest that, in situations in which 12-month-olds are required to rely on their word segmentation abilities, they give evidence of observing lexical viability constraints in the way that they parse fluent speech.     

Effect of direction of sequential presentation and redundancy on short-term recognition memory

Short-term recognition memory was tested by presenting six letters, one after the other, followed by a target letter and having S indicate whether or not the target matched one of the six letters. Recognition memory for a letter was better when it was embedded in a six-letter word, rather than a nonword, and when it was included in a sequence presented left-to-right, rather than right-to-left (Experiment 1). Reducing the presentation rate from 4/sec to 2.5/sec largely eliminated the left-to-right effect (Experiment 2). The effect of direction of presentation was greater for redundant (Experiment 1) than for nonredundant sequences (Experiment 3) and was greater for Ss who more frequently formed a word out of the sequence (Experiments 1 and 2), but was no greater for words than nonwords (Experiments 1 and 2) and no greater for letter than for line-figure sequences (Experiment 3). These findings suggest that the left-to-right effect depends as much, or more, on “peripheral” processes (e.g., eye movements) as on “central” processes (e.g., reading).     

Neighborhood effects in visual word recognition: Facilitatory or inhibitory?

In five experiments, in which subjects were to identify a target word as it was gradually clarified, we manipulated the target's frequency of occurrence in the language and its neighborhood size—the number of words that can be constructed from a target word by changing one letter, while preserving letter position. In Experiments 1–4, visual identification performance to screen-fragmented words was measured. In Experiments 1 and 2, we used the ascending method of limits, whereas Experiments 3 and 4 presented a fixed-level fragment. In Experiment 1, there was no relation between overall accuracy and neighborhood size for-words-between three and six letters in length. However, more errors of commission (guesses) were made for high-neighborhood words and more errors of omission (blanks) were made for low-neighborhood words. Letter errors within guesses occurred at serial positions having many neighbors, and these positions were also likely to contain consonants rather than vowels. In Experiment 2, a smallfacilitatory effect of neighborhood size on bothhigh- and low-frequency words was found. In contrast, in Experiments 3 and 4, using the same set of words,inhibitory effects of neighborhood size, but only for low-frequency words, were found. Experiment 5, using a speeded identification task, showed results parallel to those of Experiments 3 and 4. We suggest that whether neighborhood effects are facilitatory or inhibitory depends on whether feedback allows subjects to disconfirm initial hypotheses that the target is a high-frequency neighbor.     

Attentional priming and visual search in pigeons

Advance information about a target's identity improved visual search efficiency in pigeons. Experiments 1 and 2 compared information supplied by visual cues with information supplied by trial sequences. Reaction times (RTs) were lower when visual cues signaled a single target rather than two. RTs were (Experiment 1) or accuracy improved (Experiment 2) when a sequence of trials presented a single target rather than a mixture of 2. Experiments 3, 4, and 5 considered the selectivity of visual priming by introducing probe trials that reversed the usual cue-target relationship. RT was higher following such miscues than following the usual 1- or 2- target cuing relationships (Experiment 3); the miscuing effect persisted over variations in the target's concealment (Experiments 4 and 5), but did not occur when the target was presented alone (Experiment 4). The findings indicate that priming modifies an attentional mechanism and suggest that this effect accounts for search images.     

Short-term memory for the timing of auditory and visual signals.

Short-term memory for the timing of irregular sequences of signals has been said to be more accurate when the signals are auditory than when they are visual. No support for this contention was obtained when the signals were beeps versus flashes (Experiments 1 and 3) nor when they were sets of spoken versus typewritten digits (Experiments 4 and 5). On the other hand, support was obtained both for beeps versus flashes (Experiments 2 and 5) and for repetitions of a single spoken digit versus repetitions of a single typewritten digit (Experiment 6) when the subjects silently mouthed a nominally irrelevant item during sequence presentation. Also, the timing of sequences of auditory signals, whether verbal (Experiment 7) or nonverbal (Experiments 8 and 9), was more accurately remembered when the signals within each sequence were identical. The findings are considered from a functional perspective.     

All words are not created equal: expectations about word length guide infant statistical learning

Infants have been described as ‘statistical learners’ capable of extracting structure (such as words) from patterned input (such as language). Here, we investigated whether prior knowledge influences how infants track transitional probabilities in word segmentation tasks. Are infants biased by prior experience when engaging in sequential statistical learning? In a laboratory simulation of learning across time, we exposed 9- and 10-month-old infants to a list of either disyllabic or trisyllabic nonsense words, followed by a pause-free speech stream composed of a different set of disyllabic or trisyllabic nonsense words. Listening times revealed successful segmentation of words from fluent speech only when words were uniformly disyllabic or trisyllabic throughout both phases of the experiment. Hearing trisyllabic words during the pre-exposure phase derailed infants’ abilities to segment speech into disyllabic words, and vice versa. We conclude that prior knowledge about word length equips infants with perceptual expectations that facilitate efficient processing of subsequent language input.     

词素位置概率在中文阅读中的作用：词汇判断和眼动研究

中文词切分的认知机制一直是心理语言学关注的焦点问题之一,研究发现中文读者可使用词素位置概率等语言学线索帮助词切分,而首、尾词素位置概率的重要程度以往研究观点并不一致。本研究通过词汇判断和眼动实验,考察中文读者对首、尾词素位置概率的利用情况。实验1a和1b采用词汇判断任务,考察在不同词频条件下词首与词尾词素位置概率对词汇识别的影响。在错误率和反应时指标上,高频词条件下词首与词尾词素位置概率效应不显著;低频词条件下词首词素位置概率效应显著,词尾词素位置概率效应不显著。实验2a和2b采用句子阅读任务,考察在自然阅读情境中被试对词素位置概率的运用。在凝视时间、回视路径时间和总注视时间指标上,低频词条件下词首词素位置概率效应显著,词尾词素位置概率效应不显著。高频词条件下词首与词尾词素位置概率效应均不显著。词汇判断和眼动证据共同表明,词素位置概率信息是中文读者重要的语言词切分线索,且与词尾词素位置概率相比,词首词素位置概率在词汇切分与识别过程中发挥的作用更大。同时,词素位置概率线索的运用会受到词频的影响,研究结果支持复合词加工的混合通达表征模型。     

Are all letters really processed equally and in parallel? Further evidence of a robust first letter advantage

This present study examined accuracy and response latency of letter processing as a function of position within a horizontal array. In a series of 4 Experiments, target-strings were briefly (33 ms for Experiments 1 to 3, 83 ms for Experiment 4) displayed and both forward and backward masked. Participants then made a two alternative forced choice. The two alternative responses differed just in one element of the string, and position of mismatch was systematically manipulated. In Experiment 1, words of different lengths (from 3 to 6 letters) were presented in separate blocks. Across different lengths, there was a robust advantage in performance when the alternative response was different for the letter occurring at the first position, compared to when the difference occurred at any other position. Experiment 2 replicated this finding with the same materials used in Experiment 1, but with words of different lengths randomly intermixed within blocks. Experiment 3 provided evidence of the first position advantage with legal nonwords and strings of consonants, but did not provide any first position advantage for non-alphabetic symbols. The lack of a first position advantage for symbols was replicated in Experiment 4, where target-strings were displayed for a longer duration (83 ms). Taken together these results suggest that the first position advantage is a phenomenon that occurs specifically and selectively for letters, independent of lexical constraints. We argue that the results are consistent with models that assume a processing advantage for coding letters in the first position, and are inconsistent with the commonly held assumption in visual word recognition models that letters are equally processed in parallel independent of letter position.     

Selective learning enabled by intention to learn in sequence learning

This study investigated whether a target sequence that people intend to learn is learned selectively when it is interleaved with another (non-target) sequence. Three experiments used a serial reaction time task in which different spatial and color stimuli occurred alternately. Each of the two interleaved sequences had structural regularity. Participants in an intentional learning group were instructed to learn the target (spatial) sequence whereas those in an incidental learning group were not. In Experiments 1 and 2 spatial and color sequences were correlated. Results showed that the intentional group learned the spatial sequence better than the incidental group and learned it independently of the color sequence, whereas the incidental group learned the two sequences as a combined sequence. In Experiment 3 the sequences were uncorrelated. Results showed that the intentional group was no longer superior in learning the spatial sequence. Findings indicate that the intention to learn a target sequence enables selective learning of it only when it is correlated with a non-target sequence.     

Distribution of Programming in a Rapid Aimed Sequential Movement

Studies indicate that rapid sequential movements are preprogrammed and that preprogramming increases with complexity, but more complex sequences that require on-line programming have seldom been studied. The purpose of this investigation was to determine whether on-line programming occurs in a 7-target sequence in which there is a unique target constraint and if so, to determine how different task constraints affect the distribution of additional programming. Subjects contacted seven targets with a hand-held stylus as quickly as possible while maintaining a 90% hit rate. Initiation- and execution-timing patterns and movement kinematics were measured to determine when the additional programming took place. Results indicated that additional programming occurred before initiation and during movement to the first target when the constraint required more spatial accuracy (small target). A different type of unique target (a triple hit of one target) caused the additional programming to occur on-line one or two segments before its execution. Different positions of the unique target also affected timing patterns. Results were discussed in terms of: (1) capacity of processing; (2) control of movement variance; and (3) mean velocity as a programmed parameter in sequential aiming movements.     

Cognitive operations in the generation effect on a recall test: role of aging and divided attention

Generation effect (generated words are better memorized than read words) of anagrams, rhymes, and associates of target words was examined in young, elderly, and very old subjects. Experiments 1 and 2 showed that only young subjects benefit from the generation effect in a free-recall test when the rule is of a phonological nature. Experiments 3, 4, and 5 showed that the generation effect of rhymes was due to a resources-dependent self-initiated process. Experiments 4 and 5 showed that in a divided-attention situation, generation effect of rhymes is not significant in young subjects, but that the generation effect of semantic associates remains significant for both groups (Experiment 5). The results are discussed within the environmental support framework and the transfer-appropriate processing framework.     

Recurrent speech patterns as cues to the segmentation of multisyllabic sequences.

First and second language acquisition both require that speech be segmented into familiar, multiphonemic units (e.g., words and common phrases). The present research examines one segmentation cue that is of considerable theoretical interest: the repetition of fixed sequences of speech. On each trial, subjects heard repetitions ('pre-exposures') of two artificially-constructed, multisyllabic patterns that shared an embedded segment 1 or 2 syllables long (e.g., 2 shared syllables: [ga-li-SE] and [li-SE-stu]). There were 2 and 6, 4 and 4, or 6 and 2 repetitions of the two patterns, randomly ordered. Subjects were then to indicate the groupings they perceived within a subsequent, longer sequence containing both of the pre-exposed patterns (e.g., [ga-li-SE-stu]). Responses varied systematically with the size of the embedded segment, the repetition frequencies of the two pre-exposed patterns, and the serial position of each pre-exposure. The results illustrate how investigations of the processing of speech patterns may contribute to an understanding of some elementary aspects of language learning.     

Implicit learning of predictive relationships in three-element visual sequences by young and old adults

Knowledge of sequential relationships enables future events to be anticipated and processed efficiently. Research with the serial reaction time task (SRTT) has shown that sequence learning often occurs implicitly without effort or awareness. Here, the authors report 4 experiments that use a triplet-learning task (TLT) to investigate sequence learning in young and older adults. In the TLT, people respond only to the last target event in a series of discrete, 3-event sequences or triplets. Target predictability is manipulated by varying the triplet frequency (joint probability) and/or the statistical relationships (conditional probabilities) among events within the triplets. Results reveal that both groups learned, though older adults showed less learning of both joint and conditional probabilities. Young people used the statistical information in both cues, but older adults relied primarily on information in the 2nd cue alone. The authors conclude that the TLT complements and extends the SRTT and other tasks by offering flexibility in the kinds of sequential statistical regularities that may be studied as well as by controlling event timing and eliminating motor response sequencing.     

Further evidence of the intricacy of memory span

Memory span was measured for lists of verbal items constructed such that the items in the first half of the list were of one category and those of the second half were of another. In Experiment 1, the lists consisted of digits and words (e.g., 2, 8, 7, horse, cow, sheep or horse, cow, sheep, 2, 8, 7); in Experiment 2, they consisted of words from the same semantic domain and words from different semantic domains; in Experiments 3 and 4, they consisted of words that rhymed and words that did not rhyme. A category-order effect occurred in each experiment: Span was larger when the digits, same-domain words, or rhyming words occurred in the first half of the list than when they occurred in the second half. These findings suggest that memory span is more complex than is generally assumed.     

The role of physical and conceptual properties in preserving object continuity

Six experiments investigated the nature of the object-file representation supporting object continuity. Participants viewed preview displays consisting of 2 stimuli (either line drawings or words) presented within square frames, followed by a target display consisting of a single stimulus (either a word or a picture) presented within 1 of the frames. The relationship between the target and preview stimuli was manipulated. The first 2 experiments found that participants responded more quickly when the target was identical to the preview stimulus in the same frame (object-specific priming). In Experiments 3, 4, 5, and 6, the physical form of the target stimulus (a word or picture in 1 frame) was changed completely from that of either preview stimulus (pictures or words in both frames). Despite this physical change, object-specific priming was observed. It is suggested that object files encode postcategorical information, rather than precise physical information.     

Sources of sentence constraint on lexical ambiguity resolution

Results from a series of naming experiments demonstrated that major lexical categories of simple sentences can provide sources of constraint on the interpretation of ambiguous words (homonyms). Manipulation of verb (Experiment 1) or subject noun (Experiment 2) specificity produced contexts that were empirically rated as being strongly biased or ambiguous. Priming was demonstrated for target words related to both senses of a homonym following ambiguous sentences, but only contextually appropriate target words were primed following strongly biased dominant or subordinate sentences. Experiment 3 showed an increase in the magnitude of priming when multiple constraints on activation converged. Experiments 4 and 5 eliminated combinatorial intralexical priming as an alternative explanation. Instead, it was demonstrated that each constraint was influential only insofar as it contributed to the overall semantic representation of the sentence. When the multiple sources of constraint were retained but the sentence-level representation was changed (Experiment 4) or eliminated (Experiment 5), the results of Experiments 1, 2, and 3 and were not replicated. Experiment 6 examined the issue of homonym exposure duration by using an 80-msec stimulus onset asynchrony. The results replicated the previous experiments. The overall evidence indicates that a sentence context can be made strongly and immediately constraining by the inclusion of specific fillers for salient lexical categories. The results are discussed within a constraint-based, context-sensitive model of lexical ambiguity resolution.     

A computational model of word segmentation from continuous speech using transitional probabilities of atomic acoustic events

Word segmentation from continuous speech is a difficult task that is faced by human infants when they start to learn their native language. Several studies indicate that infants might use several different cues to solve this problem, including intonation, linguistic stress, and transitional probabilities between subsequent speech sounds. In this work, a computational model for word segmentation and learning of primitive lexical items from continuous speech is presented. The model does not utilize any a priori linguistic or phonemic knowledge such as phones, phonemes or articulatory gestures, but computes transitional probabilities between atomic acoustic events in order to detect recurring patterns in speech. Experiments with the model show that word segmentation is possible without any knowledge of linguistically relevant structures, and that the learned ungrounded word models show a relatively high selectivity towards specific words or frequently co-occurring combinations of short words.