Keyboarding

The notes I handle no better than many pianists. But the pauses between the notes - ah, that is where the art resides.
    -- Artur Schnabel
  1. Introduction
    1. Typing 90 words per minute @ 5 characters + 1 space per word = 450 keystrokes per minute, 1 keystrokes per 133 msec
    2. Piano, with 3 notes per hand @ 3 notes per second = 18 keystrokes per second, 1 keystroke per 56 msec
    3. Time to respond to sensory stimulation is 200 ms
    4. Therefore such rapid movements must be preprogrammed
  2. Reaction time
    1. Simple reaction time
      1. one signal and one response
      2. faster
      3. shorter RTs if younger, alert, practiced, can predict when the signal will occur, if the signal of brighter or louder
    2. Choice reaction time
      1. multiple signals associated with multiple responses
      2. slower unless probability of one stimulus and response is higher
      3. Faster RTs for more likely stimuli may be due to stimulus processing, response preparation or both
        1. Experiment - map more than one stimulus to a response [S1, S2 R1 & S3, S4 R2]
        2. If response preparation, choice RTs reduced by repeated testing of response regardless of stimulus
        3. If stimulus processing, choice RTs reduced by repeated testing with particular stimulus. Switching to another stimulus should yield long RTs whether the same or a different response is used
        4. Results - both response preparation and stimulus processing play a role. When S1 repeatedly, R1 RT . Then when S1 replaced with S2, R1 RT increased some but not as much as when S1 replaced with S3 or S4.
      4. Choice RT increases with the number of choices, i.e., S-R pairs
        1. Hick-Hyman law, choice RT increases linearly with log2 S-R pairs
          1. Decision making characterized by bits needed to uniquely define particular response
          2. Slope of line indicates rate at which decisions are made
          3. Interpretation in terms of information theory (Shannon) like Fitts did
          4. Set back when shown that amount of information held in short-term memory was related to how many meaningful chunks it included rathter than just 'bits'
    3. Stimulus-response compatibility
      1. If the response to a stimulus is easily mapped onto the stimulus, then RT is faster and less perturbed.
      2. Experiment - finger on vibrator, press it when it vibrated. When more vibrators & fingers used, there was no increase in RT with increasing choices when stimulus and response were compatible. When an arbitrary finger was used in response to a given vibration, then the typical increasing RT with # choices was seen.
    4. Response-response compatibility
      1. Interactions between responses so that the choice RT for a given response varies according to the other responses that are tested
      2. Experiment - choice RT with 2 finger presses. In one condition the 2 fingers were right index and right middle. In the other condition the 2 fingers were right index and left index.
      3. Results - choice RT from right index was shorter when left index finger was other response
      4. Explained by independence of left & right index fingers and linkage of right index and right middle fingers
        1. movements of middle fingers accompanied by index finger movements
  3. Sequences of finger presses
    1. Simultaneous keypresses - chords
      1. Little research on this topic
      2. Occurs not only in piano but also courtroom stenographers and typing Japanese characters
      3. Rabbitt et al. (1975) Measure RT to produce arbitrary chords
        1. choice RT increase with number of comprising keystrokes
        2. choice RT shorter for responses using homologous fingers of two hands
        3. choice RT also influenced by number of keystrokes in preceding chords. the more keystrokes in preceding chord, the longer the RT for the subsequenct chord
      4. Gopher et al. (1985) - measured chord performance (speed and accuracy) in different conditions
        1. Letters indicated particular chord. Chords for the two hands organized in three ways to dissociate spatial and anatomical mapping
          1. spatial congruence - position of keys matched for the two hands so that different fingers used
          2. manual congruence (hand symmetry) - homologous fingers used
          3. combine spatial & manual congruence - vertical keyboard
        2. Performance best in under combined spatial and manual congruence, worse under spatial congruence and worst under manual congruence
      5. What else would you like to know? Where should investigations in this area go?
    2. Sequences of single keypresses
      1. serial choice RT task - successive keystrokes elicited by series of signals
      2. Rabbitt & Vyas (1970)
        1. use index & middle finger of either hand, guided by 1 of 4 digits on screen, presented in random pattern
        2. Ss fastest when current finger was homologous to finger used on other hand in response to previous signal
        3. Ss slower when current finger on same hand as used to respond to previous signal
        4. slowest when current finger nonhomologous on opposite hand to that used to respond to previous signal
        5. Interpretation that execution of movement with one finger delays preparation of movement of other fingers on that hand. Preparation of movements of fingers on different hands can proceed independently and thus faster...
        6. homologous-finger effect possibly due to functional linkages between neural representations of homologous fingers
      3. Performance of structured stimulus sequences
        1. when patterns become predictable, performance becomes faster and more accurate
        2. surmise that representation of sequence of movements is more than an element-to-element chain of associations - somehow context and rules are encoded
        3. such a view entails hierarchical representation in which rules are induced from instances
      4. Rosenbaum et al. (1983) - Are keypress sequences executed in hierarchical manner?
        1. Ss performed keyboard sequences from memory, using left & right index & middle finger, e.g., I i I i M m M m
        2. perform sequence 6 times quickly and accurately
        3. Latency and errors varied systematically with movement position in sequence
          1. First keypress of each subsequence slower than subsequent
        4. Results accounted for using tree-traversal model of retrieval
          1. e.g., i following I is faster because homologous finger effect, stay in the index finger subroutine
          2. next I somewhat slower because of need to go up another node
          3. M following i slower again because of need to go to top of tree before calling middle finger subroutine to call right middle finger subroutine
          4. Each retrieval takes additional time and introduces possibility of error
    3. Control of rhythm and timing
      1. Now have one finger hitting one key at particular pace...
      2. Ss press key at rate specified by metronome. Data collected when metronome turned off. How well can Ss keep a steady pace?
      3. Two main, consistent findings
        1. Times between successive keypresses are negatively correlated; if one is long subsequent one is short & vice versa
        2. variance of intervals between keypresses increases with the mean of the interval
      4. Wing & Kristofferson (1973) timing model
        1. The trigger signal for each movement is separated by interval C.
        2. The keypress (R) is generated after a motor delay (D).
        3. The interval (I) between successive keypresses is Ci + Di - Di-1
        4. Thus, if Di-1 happens to be short, then the interval is longer. If Di-1 happens to be long, then the interval is shorter. This is the observed negative correlation.
        5. Variance of keypress intervals increases with size of interval because C is a random variable. As it increases, the size of its variations increases
        6. But the motor delay variance does not change with average keypress interval
        7. Parkinson's patient with increased clock variance but unchanged motor delay - dissociation indicates separate neural pathways responsible for pacemaking and for generating keypress movement
    4. Hierarchical timekeepers
      1. The previous model will not 'keep the beat' because it is not hierarchical
      2. e.g., consider drummer keeping the beat on the bass drum while hitting other drums at varying tempos
      3. Vorberg & Hambuch model
        1. postulate high-level timers that control major beat intervals and lower-level timers that control pace of other movements
        2. By measuring variance of timed movements under different conditions, various hypotheses can be tested
          1. e.g., during piano playing the variance of downbeat delays was less than the sum of the variances of the individual notes between downbeats. This finding consistent for performance of sequences with unequal intervals
          2. however, performance of sequences with equal intervals does not show this independence of variances
      4. Different rhythmic movement patterns are accomplished with different strategies
    5. The amodality of timing
      1. relation of movement timing to timing perceptual events
      2. Compare performance of movement & perceptual timing, Keele et al. (1985)
        1. variance in finger and foot tapping correlated within a subject and variable across Ss
        2. variance in perception of timing of auditory stimulus sequences correlated with rhythmic movement timing variance
        3. Surmise that one brain structure keeps time for both movement and perception
      3. Difficult to control production of one rhythm when another rhythm produced with other hand or perceived, heard
      4. Investigate preparation for timing production or perception, Rosenbaum & Patashnik, 1980
        1. Production task
          1. press left index followed by right index after specific interval
          2. measure RT to make first left index press
          3. RT decreased with increasing interval, lowest when right finger not used at all
            1. Stringent and relaxed accuracty conditions applied
            2. Slower with stringent accuracy required
            3. Faster when less accuracry required
          4. Variance increased with increasing interval
            1. less variance when more accuracy enforced
        2. Perceptual judgment task
          1. press left index finger followed by vibration of right index finger
          2. Ss estimate interval until vibration stimulus. used the same intervals as in production task
          3. measure RT to make first index press
          4. RT decreased with increasing interval to estimate
        3. The pattern of RTs in the production task matched what was seen in perceptual judgment task
        4. This indicates that preparation for timing production or judgement may not be specific to one system
    6. Integration of serial order and timing
      1. relative timing specifies serial order...
      2. so serial order and timing may not be produced by separate mechanisms
      3. investigation of coordination of serial order and timing, Summers (1975)
        1. Ss press 9 keys in specific order and rhythm given by flashing lights. Produce the same order but with 2 different rhythms
          1. rhythm 1: 500-500-100 ms
          2. rhythm 2: 500-100-100 ms
        2. perform sequence for 500 trials to become well practiced
        3. for test, have Ss reproduce sequence as fast as possible from memory
        4. the sequence was produced faster, but the rhythm corresponded to what was learned
        5. interpret that the memory representation of the sequence included relative durations
    7. Adjusting the rate of production for entire sequences
      1. Terzuolo & Viviani (1980) finding of changes in overall sequence duration produced with unchanging individual movement intervals
        1. corresponds to the handwriting result
        2. propose that timing of overall movement done by multiplicative factor
      2. Gentner (1987) reconsidered hypothesis and analyzed data using more rigorous statistical methods and did not support predictions
  4. Typewriting
    1. Historical issues
      1. typewriter invented in mid 1800s
      2. arrangement of keys is not optimal but is accident of history
      3. central issue investigated early on was relation between gaze and typing
        1. eye leads the hand by 4-8 letters, eye-hand span
        2. pattern of gaze over text during typing is different from that observed during reading for comprehension. eye movements organized to specifically provide letter information at rate needed for typing
      4. typing words in random order is as fast as typing words in correct order
      5. typing words is faster than typing nonwords
        1. manipulate amount of preview available to typist
          1. normal prose typed at fastest level with as little as 8 characters preview
          2. random letter strings take longer to type even with 8 letter preview
    2. Units of typing control
      1. because words are typed more quickly than nonwords, and word order does not affect speed, the production unit during typing is no larger than one word
      2. experiment - have typist stop typing when a tone sounds, Logan, 1982.
        1. if words were produced as a whole, then remaining characters of a given word should be typed after the tone
        2. but typist stop within as easily as between words
      3. typists also stop within words to correct errors, often as soon as the error was typed
        1. some incorrect keystrokes are of less force and longer latency than correct keystrokes [if the brain is uncertain, why is the movement produced at all?]
      4. nature of errors
        1. most frequent, horizontal adjacent key, e.g., midtake for mistake
        2. next most frequent, vertical adjacent key, e.g., mixtake for mistake
        3. third most frequent, homologous finger, e.g, miltake for mistake
        4. but when wrong key is hit, it is with the appropriate finger for that key
          1. thus finger selection is probably a stage of preparation
    3. Timing of keystrokes in typewriting
      1. high frequency words in a language typed faster than low frequency words
      2. high frequency digraphs (2 letter combinations) typed faster than low frequency digraphs
      3. keystroke timing also depends on relations between fingers and hand making strokes
        1. keystrokes produced by opposite hands faster than keystrokes produced by different fingers of the same hand
        2. keystrokes produced by different fingers of the same hand faster than keystrokes produced by same finger
      4. is speed of producing high frequency words & digraphs due to mechanical factors or to cognitive factors?
        1. Experiment - type strings from memory following presentation of a reaction signal that was preceded by a series of warning tones to enable Ss to prepare to respond, Sternberg et al., 1978
          1. latency of first keystroke increased with number of letters to type [this finding corresponds to what we saw in serial movement tasks]
          2. rate of increase of greater for letters with alternating hands than with single hand
          3. keystroke duration increased with number of letters, more for within-hand than between-hand typing.
          4. keystroke times for nonwords were the same as for words
          5. Interpretation -
            1. when adequate preparation is allowed, differences persist for within-hand versus between-hand typing, so these are due to mechanical factors during execution
            2. nonword vs. word difference not due to execution because it vanished when Ss could be prepared
        2. Experiment - determine why high frequency words & digraphs are typed faster, Gentner et al., 1988
          1. measure speed of typists in different languages
            1. variations of digraph speed were the same for American & Dutch typists
            2. Since digraph frequency was different in the two languages, digraph effects must be mechanical in origin
            3. variations in word speed did vary with language, thus, of cognitive origin
            4. also, practice improved speed of low frequency but not high frequency words, but practice did not improve digraph speed
    4. Rumelhart and Norman's (1982) model of typewriting
  5. Piano playing
    1. unlike in typing, the rhythm and force of the keystroke are imperative in playing piano
    2. Experiment - Sloboda (1983)
      1. Pianist Ss played 2 scores with the same notes but different different measures
      2. Recorded time and force of keystrokes
      3. although the timing of keystrokes was the same (8.19 bottom), the force of the keystrokes varied systematically (8.19 top), being shifted one note
    3. pianists can control timing and force independently and so communicate musical meter by force alone