|
NEW ENGLAND SEQUENCING AND TIMING (NEST)
Eighteenth Annual Meeting
Time: Saturday, March 8, 2008, 8:30 a.m. – 5:00 p.m. Place: Haskins Laboratories, 300 George Street, New Haven, CT Organizer: Bruno H. Repp (e-mail: repp@haskins.yale.edu) Assistant: Minjung Son PROGRAM 8:30 – 9:00 Continental breakfast
9:00 – 9:25
Time estimation plays a major role in many cognitive tasks, but is usually studied in relative isolation. As a consequence, many general characteristics of cognition like attention and learning have been incorporated in theories of time estimation specifically tailored to explain particular empirical findings. For example, effects of attention are often explained by an attentional gate that is part of the time estimation model. In my talk I will show that a more general theory of attention (and memory) can explain the existing data, and has made successful predictions for new experiments. The theory is explicated in a model based on the ACT-R cognitive architecture, which allows detailed predictions of empirical data. If time permits, I will briefly show some more recent results that suggest that people are capable of adding up time intervals, and that the logarithmic nature of time perception biases their estimates.
9:25 – 9:50
Memory is an integral part of Scalar Expectancy Theory (SET); however, little is known regarding how the number of intervals learned or the duration of a retention interval affects recall of time intervals. To address this question, two experiments examined reproduction of one or more time intervals after a delay of minutes to hours. First, groups of participants learned one interval, two intervals, or three intervals, ranging from 1.1 to 3.3 s, and then reproduced the intervals without feedback immediately after the training blocks, and again 1 hr after training. In the single-interval groups, temporal reproductions drifted progressively longer in subsequent testing trials, and longer intervals drifted significantly more than shorter ones. This drift in temporal reproductions in the absence of feedback was attenuated in the multiple-interval groups. To examine the possibility of anchoring effects, a second experiment manipulated the range of times learned. No range effects were found, but a significant order-of-training effect indicated that drift occurred only when the longer interval was trained last; when participants learned the longer interval first, drift was attenuated. Results will be discussed in terms of SET’s K* parameter and the leading hypotheses of the psychophysics of memory.
9:50 – 10:15 Does the ability to detect a temporally deviating interval in a sequence of sounds depend on the synchronization between these sounds and an inner oscillatory cycle? Oscillatorbased timing mechanisms predict a performance decrease if the cycle of the inner oscillatory process is not in phase with the outer time events. Pacemaker-counter models on the other hand assume no influence of phase shifts on performance. Previous studies used sequences with empty intervals to investigate the influence of phase shifts. The results are mixed (Schulze, 1978; Keele et al., 1989; Ivry & Hazeltine, 1995; McAuley & Kidd, 1998; Pashler, 2001; Berens & Pastore, 2005). We tried to find more evidence for one of the models by using sequences with filled intervals. We compared the data from two experiments in which participants heard a sequence of seven filled intervals. Six of them had a fixed (standard) duration of either 50, 150 or 250 ms. The duration of one (comparison) interval was variable (it was lengthened by using an adaptive staircase procedure to estimate the threshold where the participants detected the deviating interval in 75% of the trials). In Experiment 1, the silent interstimulus intervals (ISIs) matched the duration of the standard intervals, while in Experiment 2 the ISI was set to 125 ms for all three of the standard durations. Whereas in Experiment 2 phase corrections are necessary to become synchronized with the onset of each interval (if the oscillator adapts its period to the time between the onsets and offsets of the filled intervals), in Experiment 1 no phase corrections were required. If the oscillator adapts its period only to the duration between onsets of the filled intervals, no phase corrections are required but performance will differ between the experiments according to Weber’s law. Our results show no performance differences between the two experiments. Thus, our results question the assumptions of an oscillator-based model and favor a pacemaker-counter mechanism. 10:15 – 10:45 Coffee break 10:45 – 11:10 CONTEXTUAL EFFECTS ON THE PERCEPTION OF DURATION IN SPEECH AND NON-SPEECH John Kingston1, Shigeto Kawahara2, Della Chambless3, Daniel Mash1, and Eve Brenner-Alsop1 (1University of Massachusetts; 2University of Georgia; 3University of NorthTexas) In this talk, we report the results of experiments designed to test the competing predictions of direct realist vs auditorist models and of autonomous vs interactive models of speech perception. We compared Japanese, Norwegian, Italian, and English listeners’ identification and discrimination of speech stimuli in which the durations of a vowel and a following consonant were varied orthogonally. Listeners with the same language backgrounds also identified non-speech analogues constructed by replacing the vowels with anharmonic sine wave complexes. Vowel duration covaries directly with following consonant duration in Japanese but inversely in Norwegian, Italian, and English, so these experiments test the top-down application of linguistic experience in these tasks. The non-speech analogues were used to test the hypothesis that durational contrast transforms the acoustic durations of the vowel and consonant intervals (or their analogues) as they pass through the auditory system. Japanese listeners identified the consonant intervals in the speech stimuli as long more often after longer vowels, while Norwegian, and Italian listeners instead did so more often after short vowels. These biases reflect these listeners’ linguistic experience. English listeners resembled Japanese listeners despite the fact that vowel and following consonant durations covary inversely in English pronunciation. Listeners’ identification of the non-speech stimuli and their discrimination of both the speech and non-speech stimuli did not differ as a function of their linguistic experience, but instead appear to reflect their having added the durations of the two intervals in responding to these stimuli and tasks. We will also argue that the addition of the durations of successive intervals is a post-perceptual rather than auditory process. These findings provide no evidence that the percept of the duration of an interval contrasts with that of a neighboring interval and also indicate that the scope of the top-down application of linguistic knowledge is limited. They thus provide no direct support for an auditorist model of speech perception, but do support an autonomous model of speech perception over an interactive one.
11:10 – 11:35
Selective attention provides a mechanism by which people preferentially process subsets of stimuli when faced with overwhelming amounts of information. Spatially selective attention is important for perceiving complex scenes in which multiple objects are presented simultaneously at different locations. Accumulating evidence indicates that temporally selective attention plays an important role in perception when more information than can be processed in detail is presented rapidly. Specifically, it affects both auditory and visual evoked potentials at early perceptual stages of processing. However, temporally selective attention can only be considered an important perceptual tool if people employ it without instruction to process relevant, rapidly changing information such as natural speech. I will present evidence that listeners do use temporal selection to preferentially process the segments of speech they are least able to predict: the initial portions of words. These results will be discussed in the context of more general event perception and segmentation as well.
11:35 – 12:00
Most people find it easy to perform rhythmic movements in synchrony with music, which reflects their ability to extract the underlying metrical structure. By providing a temporal perceptual framework, metrical structure directs attention to critical time points and enhances the perception and performance of acoustic events at these time points (Large & Palmer, 2002). In other words, performance accuracy may reflect the effective allocation of attention to the rhythm. To investigate the allocation of attention to grouping and metrical accents in the metrical structure, and to reveal the effect of musical expertise on these processes, musicians and nonmusicians were tested in a click localization paradigm, which has been shown to reveal perceptual grouping and subjective organization in both music and speech materials. Participants listened to auditory rhythms containing a single superimposed click and localized it on a graphical representation of the rhythm. Most of the clicks occurred in or near a critical perceptual group of three tones. The positions of metrically strong and weak beats within this perceptual group were manipulated by means of preceding context. Both grouping and metrical influences on click localization were demonstrated. Moreover, whereas all participants located a click accurately at the beginning of a perceptual group, only musicians located it well when it coincided with a metrically strong beat at the end of the perceptual group. These results imply that musical expertise helps allocating attention toward the embedded metrical structure by enhancing sensitivity to both metrical and grouping accents. 12:00 – 1:00 Lunch (provided) 1:00 – 1:25 When we speak to one another, we adapt and coordinate our facial expressions in response to our inner states and in response to our perceptions of the person with whom we are speaking. We say we are speaking "with" someone, since this is a shared experience; one that creates an interpersonal coupled system with feedback between the dynamics of the two speakers' perceptions and actions. There is evidence that structural appearance of faces as represented in static photographs and the dynamics of faces may involve separate visual processing. Dissociating the contribution of facial appearance and facial dynamics to the adaptive coupled system involved in conversation requires being able to assign appearance to dynamics in real time. This talk presents a method for assignment of facial appearance to facial dynamics using Active Appearance Models in a way that is sufficiently convincing that naive conversants do not realize that they are speaking with an appearance assigned face and that is sufficiently efficient to be computed within one video frame. 1:25 – 1:50 Prior research indicates that synchronized tapping performance is far worse with flashing visual stimuli than with auditory stimuli. This difference may reflect an auditory advantage for processing temporal information, while visual processing may have the advantage with spatial information. Two finger-tapping experiments explore whether adding a spatial component, compatible or incompatible with the tapping action, can improve visuomotor synchronization performance over purely temporal flashing stimuli. In Experiment 1, synchronization success rates increased dramatically for spatiotemporal sequences of both geometric and biological forms compared to flash sequences. In Experiment 2, synchronization performance was best when target sequences and movements were directionally compatible (i.e., when both moved down simultaneously), followed by action-neutral stimuli (i.e., leftward moving target/downward tapping movement), and was poorest for incompatible moving stimuli (upward moving target/downward tapping movement) and flashing target sequences. The facilitative effects of compatible motion suggest tight perception-action linkages, and that the modality differences in sensorimotor synchronization are not as extreme as previously indicated. Additionally, preliminary results indicating that interpersonal synchrony promotes interpersonal affiliation will be reported. 1:50 – 2:15 During the past nine years the case has been presented that timing processes in tapping tasks are distinct and separable from timing processes in circle drawing tasks. This distinction has been at the center of the hypothesis that there are two kinds of timing: event-based and emergent (Ivry, Spencer, Zelaznik, & Diederichsen, 2002). Repp (2000) has demonstrated that small as well as large perturbations in the duration of one metronome interval during synchronization are corrected for rapidly in tapping. This phase correction process is a characteristic of event timing. We report three experiments that examined whether synchronization performance in circle drawing exhibits phasecorrection behavior. Participants synchronized to a metronome and then one interval was perturbed. The location, size, and direction of perturbation were not predictable. Experiment 1 showed that participants could not make phase corrections in circle drawing. However, this result was compromised by the fact that participants were not consistently synchronized in circle drawing. Experiment 2, which employed improved training and experimental procedures, did have participants synchronized prior to temporal perturbations in circle drawing. Participants did not exhibit phase corrections. In a third experiment (still ongoing), we provided a tactile anchoring event in the circle drawing task, by raising the level of the circle at the timing target. Preliminary results suggest that participants appear now to adjust for the temporal perturbation, although not in the classic phase correction manner. We infer that events, the necessary elements for event timing processes, can be inserted in circle drawing timing tasks. Such an inference, if correct, means that the event-based versus emergent timing distinction is not based solely on movement kinematics. 2:15 – 2:40 The actualization of a simple affordance task—grasping and moving wooden planks of different sizes using either one or two hands—was assessed in the context of taskrelevant (plank sequence, plank presentation speed) and task-irrelevant (cognitive load) manipulations. In Experiment 1, fast (3 sec/plank) and self-paced (~5 sec/plank) presentation speeds revealed hysteresis: The transition point for ascending series was greater than the transition point for descending series. Hysteresis was eliminated in the slowest presentation speed (10 sec/plank). In Experiment 2, hysteresis was exaggerated by a cognitive load (counting backward by seven) for both fast and slow presentation speeds. These results suggest that behavioral responses to the attractor dynamics of perceived affordances are processes that require minimal cognitive resources. 2:40 – 3:10 Coffee break 3:10 – 3:35 Studies of timing have often employed the use of a two-level hierarchical model where clock and motor components are modelled as random variables. The interaction of such stochastic processes has been studied from the framework of linear timekeeper models (Wing & Kristofferson, 1973), broadcast models (Rosenbaum, 2002), and nonlinear coupled oscillator models (Yamanishi, Kawato, & Suzuki, 1980; Schöner, 2002). All of these models make varying philosophical assumptions about the independence and interdependence of the above-mentioned random processes. Here we present a formal method called a graphical model (Jordan & Ghahramani, 2004) that can identify the above-mentioned dependencies from experimentally obtained data. A graphical model is a family of probability distributions defined in terms of a directed and/or undirected graph. The nodes in the graph are identified with random variables, and joint probability distributions are defined by taking products over functions that are defined on connected subsets of nodes. This formalism provides the ability to compute marginal and conditional probabilities of interest, so one may test the performance of a kind of graph architecture that includes hierarchical latent variables against experimentally obtained data. Although the graphical model itself is agnostic to the distinction between frequentist, Markovian, and Bayesian statistics, such models are often used as general Bayesian inference engines. In the present research, ten healthy subjects took part in an experiment where they had to tap their right index finger (synchronization-continuation) in intervals of 500 ms and 1000 ms. On the basis of the experimentally obtained timing errors, two possible dependencies were tested using graphical model architectures: 1) clock and motor processes are independent and 2) clock and motor processes have latent dependencies. The data strongly indicate that there are multiple dependencies between the clock and motor processes, thus supporting the idea that the characteristics of the motor system might influence the timer in non-trivial ways. 3:35 – 4:00 When people listen to music, they often move their body in time with the beat. However, people differ widely in their ability to ‘feel a beat’. Why? Here we combined functional magnetic resonance imaging with a timing task that is diagnostic of individual differences in beat perception, and compared the brain activity of individuals who readily perceive an implied beat with those who do not. The groups show different activation in auditory and motor areas, even when no behavioral differences occur. The results support two conclusions. First, beat perception is mediated by the activation of motor timing circuits involved in rhythm production. Second, some individuals more readily engage these beatbased circuits when making timing judgments than do others. 4:00 – 4:25 Waiting requires suppressing responses until the right time or the right stimulus has occurred. Here, we describe a neural system in rats that controls waiting during a delayed response task. First, we show that inactivating rodent prefrontal cortex impairs rodents' ability to wait in tasks with variable and fixed delays. Then, we show that inactivating rodent prefrontal cortex specifically decreases delay-related neural activity in rodent motor cortex. Finally, we show that neurons in rodent prefrontal cortex are modulated while animals wait for rewards and fire differently depending on recent reward history. These data provide strong evidence that in rats, prefrontal regions are critical to the temporal control of action. 4:25 – 4:50 A common assumption in models of motor control is that positioning movements are made by moving from one position to another. The assumption is supported by evidence consistent with the equilibrium point model, by data indicating knowledge of future positions before movements to those positions are initiated, and by the success of computational models of motor control, such as the posture-based (PB) motion planning model of my colleagues and me, which suppose that goal positions have special status in the planning of forthcoming movements. Such models assume that the motor system can be in either of two states: holding still or moving. In this talk, I question this assumption. I consider the alternative hypothesis that as long as we are alive, we are always moving. This hypothesis is supported by everyday observations about breathing, heartbeat, and other basic functions. Moreover, and more subtly, it turns out that assuming a single state for living animals – being always in motion – allows for the beginnings of a more coherent, unified theory of motor control than may be allowed by assuming a division between stasis and motion. Disparate findings, including new findings from my lab, are accommodated by this new perspective. 5:00 – 6:20 Drinks at Bar (254 Crown Street) 6:30 – 8:30 Dinner at Bentara (76 Orange Street) |