Tenth Annual Meeting

Time:              Saturday, March 4, 2000, 8:30 a.m. - 4:30 p.m.

Place:              Lecture Room, Sterling Memorial Library, Yale University

                        120 High Street (entrance on Wall Street), New Haven, CT

Organizer:     Bruno H. Repp, Haskins Laboratories

                        E-mail: repp@tom.haskins.yale.edu

8:30 - 9:00       Welcome (coffee, tea, Danish, muffins)

9:00 - 9:30

PAVLOVIAN CONDITIONING INTERSTIMULUS INTERVAL FUNCTION EMERGES FROM SYNAPTIC NOISE AND NEURONAL DYNAMICS

Thomas H. Brown, John P. McGann, and Michael P. McCreless

Department of Psychology and Interdepartmental Neuroscience Program,

Yale University

E-mail: thomas.brown@yale.edu

We created a neurobiologically-inspired, real-time learning model that accounts for temporal encoding phenomena observed in Pavlovian conditioning. The 2000 neuron circuit was previously shown to encode accurately the interstimulus interval (ISI), the time interval between the onset of the conditioned stimulus (CS) and the onset of the unconditioned stimulus (US). ISIs are easily learned in the range of tenths of a second to tens of seconds. The introduction of synaptic noise into the model causes the accuracy of temporal encoding to follow a Weber-like law, in that the temporal variability of the conditioned response is proportional to its peak latency.  One time-domain difficulty that other conditioning models have experienced concerns what can be called the ISI function: the relationship between the CS-US interval and the conditioning effectiveness. The speed or extent of conditioning generally peaks at short CS-US intervals and then declines monotonically as the ISI increases. Here we show that the ISI function can emerge naturally from Weber-like temporal encoding combined with realistic neuronal dynamics. The key dynamical feature of the neurons is that the latency to fire depends jointly on the intensity of the synaptic input and intrinsic synaptic integration times.

9:30 - 10:00

TEMPORALLY SPECIFIC BLOCKING OF THE CONDITIONED EYEBLINK RESPONSE: TEST OF A COMPUTATIONAL MODEL

John W. Moore, Vanessa Castagna, and Jordan Marks

Department of Psychology, University of Massachusetts, Amherst

E-mail: jwmoore@psych.umass.edu

To what extent is Kamin blocking temporally specific? Kamin blocking involves a two-stage training protocol. In stage 1, rabbits were trained to make eyeblink CRs to an 800-ms light or tone using one of two CS--US intervals, 300 or 700 ms. Stage-1 training promotes robust CRs that occur near the time of the US: CRs trained with a 300-ms CS--US interval peak at about 300 ms after CS onset; CRs trained with a 700-ms CS--US interval peak at about 700 ms after CS onset. In stage 2, all rabbits received additional training, but now the CS consisted of two components, the light and the tone as a compound CS. Only now, there are two CS-US intervals intermixed from trial to trial, 300 and 700 ms. Without any stage-1 training, this mixture of CS--US intervals promotes bimodal (twin-peaked) CRs. Temporally specific blocking manifests itself as suppression of CRs to the added CS (the one not present in stage 1), but only within the time window appropriate for the stage-1 CS--US interval. There would be no suppression of CRs to the added CS within the time window appropriate for the other CS--US interval employed in stage 2. Temporally specific blocking would be expected from Ralph Miller's reports of temporal encoding in fear conditioning, and it is a strong prediction of Sutton and Barto's TD (CSC) model, which is capable of generating complex topographical features of eyeblink CRs (see Rosenbaum and Collyer, Timing of Behavior, MIT Press, 1998). We performed several variations of the stage-2 protocol for temporally specific blocking. Although we observed blocking of CRs to the CS added in stage 2, this blocking was not temporally specific. This finding represents a severe challenge to real-time models of conditioning in general and the TD CSC model in particular.

10:00 - 10:30

AN EXPECTED TIME ANALYSIS OF CLASSICAL CONDITIONING

Kimberly Kirkpatrick and Russell M. Church

Department of Psychology, Brown University

E-mail: Kim_Kirkpatrick@brown.edu

The expected time hypothesis is a timing account of classical conditioning that uses three principles to explain a wide range of phenomena: (1) the mean response rate during the interval is a function of the mean expected time at the time of event delivery; (2) the form of the response rate function is determined by the shape of the expectation function; and (3) when two or more expected times are simultaneously active, then there is additive combination of the response rate functions supported separately by the events. The expected time hypothesis is a parsimonious account of the role of temporal intervals in conditioning that can predict the rate and form of responding under a wide array of interval distribution forms including fixed, random, and mixed intervals that are marked by either CS or US events. The principle of summation of response rates from multiple expected times allows for the explanation of several additional phenomena observed in conditioning procedures including: duty cycle effects, trace conditioning, contingency manipulations, and responding to events that have a presumed irrelevant relationship with food. Timing accounts of conditioning phenomena may prove more successful than traditional associative views of the conditioning process.

10:30 - 11:00  

Coffee break

11:00 - 11:30

TIMING NETS FOR AUDITORY OBJECT FORMATION AND RHYTHM PERCEPTION

Peter Cariani

Eaton Peabody Laboratory, Massachusetts Eye and Ear Infirmary, Boston

E-mail: peter@epl.meei.harvard.edu

            Some further explorations of the properties of timing nets consisting of delay lines, coincidence elements, and temporally-structured inputs have been carried out. Recurrent timing nets consist of arrays of delay loops of different lengths that build up temporal expectations from previous inputs. The networks provide a mechanism by which auditory objects can be built up from coherent, recurring time patterns and separated from other objects. They provide examples of how perceptual organization might arise out of pattern-coherence rather than local features and feature-bindings. Recent simulations have indicated that the pattern-coherence strategy can operate on a frequency-by-frequency basis to segregate double vowels with different fundamentals. Recurrent timing nets have also been applied to the representation of complex rhythmic patterns to recognize dominant periodicities. In effect the networks compute running autocorrelation functions of their inputs. Spatial patterns of activity in the outputs of simple recurrent networks show slow movement across the coincidence array when there is rhythmic motion (accelerations, decelerations). We will discuss some prospects and problems for these networks. A problem with the current simple architectures is that they lose rhythmic sub-patterns if they are jittered sufficiently; another is that such architectures represent temporal intervals (metrical time) very well, but not sequences of events (temporal ordering, chaining). Lastly, we will discuss how short-term synaptic potentiation might support reverberating networks of this kind by transiently opening gates of recurrent paths that yield high correlations of current inputs with previous ones.

11:30 - 12:00

COORDINATION OF BODY SEGMENTS IN VISUALLY GUIDED POSTURAL CONTROL

Steven M. Boker, Jennifer L. Rotondo, and David R. Parker

Department of Psychology, University of Notre Dame

E-mail: sboker@nd.edu

In stationary upright stance, a variety of sensory cues may be used as input to control posture. The current experiment varies distance to a visual target while maintaining consistent input from audition and proprioception during a stationary stance task. Subjects' head, trunk and hip motion, and orientation are recorded along with center of pressure. These data are analyzed using a lagged mutual information method and a windowed cross-correlation method to examine changes in the structure of the coordination as visual information's acuity is increased. These data and analyses are designed to explore the relationship between spatial accuracy of information from the environment and the time constants relating to intersegmental coordination. As the time constants change, a change in the sequence of movements used to maintain posture may signal a bifurcation in the dynamics of postural control. Applications of these techniques to the study of age-related change in postural control will be discussed.

12:00 - 1:00

Lunch (sandwiches, fruits, sodas, coffee, tea)

1:00 - 1:30

INDIVIDUAL DIFFERENCES IN TIMING CONSISTENCY ARE CORRELATED AMONG TAPPING, INTERMITTENT CIRCLE DRAWING, AND DURATION PERCEPTION TASKS, BUT ARE NOT CORRELATED WITH CONTINUOUS CIRCLE DRAWING

Howard N. Zelaznik¹, Rebecca M. Spencer¹, and Richard B. Ivry²

¹Department of Psychology, Purdue University

²Department of Psychology, University of California at Berkeley

E-mail: Zelaznik@sla.purdue.edu

At NEST last year we presented the results of several experiments that we believed provided evidence that timing in motor skills was not attributable to a general purpose timing ability.  In those experiments, timing precision in circle drawing was not related to timing precision in tapping (Robertson et al., JEP:HP&P, 1999).  It was argued that those results were not expected if timing in motor tasks were general.  In the present two experiments, we provide evidence that at least two classes of timing processes, explicit timing and indirect timing, are needed to understand variability in timing.  We show that performance of an auditory duration discrimination task is related to timing precision in tapping and intermittent circle drawing, but is not related to timing precision in continuous circle drawing.  Furthermore, we demonstrate that the performance in duration discrimination is specific to the variability in the duration of the pause component of the intermittent circle drawing tasks.  We believe that this evidence makes it clear that timing in tapping is fundamentally different than timing in continuous drawing tasks.  Finally, our future research plans are presented.

1:30 - 2:00

TEMPORAL DRIFT IN RHYTHMIC FINGER TAPPING

Geoffrey L. Collier¹ and R. Todd Ogden²

¹Department of Psychology and Sociology, South Carolina State University

²Department of Statistics, University of South Carolina

E-mail: Rhythmpsyc@aol.com

A model for temporal drift in simple isochronous rhythmic tapping is developed. Temporal drift is the tendency, observed both in the lab and in musical performance, for tempo to drift up and down when tapping simple (or complex) rhythms. This is seen even in the most highly trained musicians. There are two reasons why temporal drift is worthy of attention. First, the dominant model for the decomposition of variance of rhythmic tapping (Wing & Kristofferson, 1973) is not valid when temporal drift is present (e.g., Vorberg & Wing, 1996). This model decomposes temporal variability into a component attributed to central or clock processes, and a component attributed to peripheral, motoric, implementation processes. This model has been widely applied, including to situations in which disease or trauma is present (e.g., Ducheck, Balota & Ferrar, 1994; Ivry & Keele, 1989; Wing 1988). Because the model is not valid when drift is present, drift has been treated as a nuisance variable (although see Madison, in press). However, the various strategies commonly employed to deal with it are statistically problematic. In addition, we view temporal drift as a fact of temporal performance worthy of attention in its own right. Accordingly, we have developed a 3-component extension of the classic Wing & Kristofferson 2-component model of rhythmic tapping. In the classic version, timing is decomposed into a motoric and a clock component. In the 3-component version, the clock component is in turn divided into a component due to temporal drift and a component not due to temporal drift (the residual). The drift component is estimated through local linear regression with a specifiable bandwidth, with simultaneous estimation of the remaining two components. Both simulation and empirical application results will be shown.

2:00 - 2:30

ADAPTATION TO SUBLIMINAL AND SUPRALIMINAL TEMPO CHANGES IN SYNCHRONIZED FINGER TAPPING

Bruno H. Repp

Haskins Laboratories, New Haven

E-mail: repp@tom.haskins.yale.edu

            In a task requiring synchronized finger tapping to isochronous auditory sequences containing step (i.e., abrupt tempo) changes, Thaut, Miller, and Schauer (Biocybernetics, 1998, 79, 241–250) found that small step changes (2% or 4% of the baseline interval) were followed by rapid adaptation of the inter-tap interval (ITI) but only by very slow changes in the asynchronies between taps and tones, whereas a larger step change (10%) elicited initial overcorrection of the ITI followed by rapid adaptation of both ITI and asynchronies. Thaut et al. interpreted these data as revealing a priority of (overt) period matching over (overt) phase matching below the level of awareness. However, from the perspective of a two-tiered model including central timekeeper period correction and peripheral phase error correction mechanisms (Mates, Biocybernetics, 1994, 70, 463–473, 475–484), the results of Thaut et al. are ambiguous with regard to a priority among these underlying processes. Two new experiments will be reported. In Experiment 1, participants synchronized with a sequence that stopped soon after a small (2%) step change and then continued tapping freely. Although the ITI adapted almost instantly to the tempo change during synchronization, the tempo of the continuation tapping adapted more slowly. This suggests that the ITI adaptation is due to phase correction, not period correction. In Experiment 2, participants synchronized with sequences containing step changes ranging from 2% to 6% and also reported on each trial whether they had detected a change. The results largely replicated those of Thaut et al. However, contrary to their conclusions, the primary mechanism below the detection threshold is considered to be phase error correction, whereas the slow adaptation of asynchronies is attributed to slow timekeeper period correction. Above the detection threshold, awareness of a tempo change seems to lead to a deliberate timekeeper period adjustment in addition to the automatic phase error correction, thus accounting for the initial ITI overshoot and the more rapid adaptation of asynchronies.

2:30 - 3:00

Coffee break

3:00 - 3:30

ESTIMATES OF SEQUENCE ACCELERATION AND DECELERATION RATE: EVIDENCE FOR SYNCHRONIZATION OF INTERNAL RHYTHMS

Amandine Penel¹, Marie Rivenez², and Carolyn Drake²

¹Haskins Laboratories, New Haven

²Laboratoire de Psychologie Expérimentale, CNRS, and Université Paris 5

In a series of three experiments, we compared the perception of sequence acceleration and deceleration. We used a method of direct estimation of rate of tempo change, asking participants to indicate on a continuous scale how much a sequence of nine sounds accelerated or decelerated. Results did not reveal any general under- or overestimation of accelerations compared to decelerations. However, at the fastest tempi within a range of tempi presented in an experimental session, accelerations were judged as being greater than decelerations for the same physical magnitude. The opposite pattern was observed for the slowest tempi within a range. These findings are consistent with the hypothesis of the synchronization of internal rhythms with events in sequences, even if their inter-onset intervals (IOIs) are irregular: At fast tempi, the internal rhythm synchronizing with events is also attracted towards an intermediate slower tempo, making an acceleration more salient than a deceleration; at slow tempi, the synchronized internal rhythm is also attracted towards an intermediate faster tempo, making a deceleration more salient than an acceleration. Estimates varied with the tempo range presented to participants, and the results supported the hypothesis of two preferred internal rhythms, one around 200-250 ms IOI and the other around 550-600 ms IOI. These findings are consistent with Collyer, Broadbent, and Church's (1992, 1994) "oscillator signature". They also support the 400 ms IOI transition between two different zones of tempi suggested by Fraisse in 1956.

3:30 - 4:00

CONTEXTUAL EFFECTS ON CATEGORICAL TIME JUDGMENTS

J. Devin McAuley¹ and Mari Riess Jones²

¹Department of Psychology, Bowling Green State University

²Department of Psychology, The Ohio State University

E-mail: mcauley@bgnet.bgsu.edu

Three experiments examined effects of extended (session) context on accuracy of categorical time judgments and loci of indifference intervals. In Experiment 1, different groups of participants compared standard and comparison inter-onset-intervals (IOIs) ranging from 200 to 800 ms in a task where the standard was reinforced by a brief local tempo on each trial. In Experiment 1, accuracy was uniformly high, but the locus of the indifference interval was indeterminate. Experiments 2 & 3 embedded the same local tempi plus standard and comparison IOIs within different session contexts that varied according to: (1) mean session rate, (2) standard deviation, (3) range, and (4) number of different local tempi within a session. Both accuracy and specific indifference interval estimates varied as a function of mean session rate and range of tempi. Best predictors of performance involved relative range (RR = range/mean rate); performance was best with low RR and worst with high RR values.

4:00 - 4:30

TIMING IN THE CENTRAL EXECUTIVE: BIDIRECTIONAL INTERFERENCE IN TEMPORAL PRODUCTION AND RANDOM NUMBER GENERATION DUAL-TASK PERFORMANCE

Scott W. Brown and C. Tigg Frieh

Department of Psychology, University of Southern Maine

E-mail: swbrown@usm.maine.edu

Some recent research suggests that timing is a function controlled by the Central Executive part of working memory. The Central Executive is a hypothetical component of short-term memory that serves as an attentional controller mechanism that is responsible for monitoring, scheduling, and coordinating ongoing behavior. The Central Executive is associated with a pool of specialized processing resources, and concurrent tasks that rely on these resources may produce mutual interference because of capacity limitations. In the present experiment, we sought evidence of bidirectional interference by pairing a timing task with an established Central Executive task. Subjects performed temporal and nontemporal tasks both separately and concurrently in a series of 2-min trials. The temporal task required subjects to generate a repeated sequence of 5-sec temporal productions by pressing a button on a computer- linked mouse device. The nontemporal task was random number generation (RNG), a task previously associated with the Central Executive. In this task, subjects were required to verbally produce a continuous series of random numbers. The subjects performed both easy (the numbers 1-10 inclusive) and difficult (the numbers 34-43 inclusive) versions of the RNG task. A comparison of single-task versus dual-task conditions showed that the RNG task interfered with timing by making temporal production responses longer and more variable. Similarly, the timing task disrupted RNG performance in the dual- task condition by making responses less random compared to the RNG-only single task condition. Mutual interference between the timing and randomization tasks implies that both tasks rely on the same set of attentional resources. These results support the idea that timing is supervised by the Central Executive.

5:00 - 6:00

Drinks in a bar or walk on Yale campus

6:00 - 8:00

Dinner at Pika Tapas restaurant (39 High Street)

NEST Home