Review of The analysis and cognition of melodic complexity: the implication-

realization model, by E. Narmour, Univ. of Chicago Press, Chicago, 1992. Music

Perception, 1995, 12 (4), pp 486-509.

Recently, chatting with a well-known composer, I mentioned that I was engaged in

reviewing Eugene Narmour’s most recent book. “Narmour ?”, he mused, “Oh yes,
doesn’t like Schenker much, does he ? Or harmony.” Such a response is not wholly
surprising in view of the level of controversy generated in the music-analytic
community by Narmour’s “Beyond Schenkerism” following its appearance in 1977.
However, to define Narmour’s views as expressed in that book in terms of what he
argued against is to do his thesis less than justice; read as an attempt to re-evaluate
the premises of the dominant music-analytic framework in the context of an
informed understanding of cognition, the book excited as many readers in the vistas
that it opened up as it dismayed in the criticisms that it raised. Perhaps, though, the
idea of Narmour as being an opponent rather than a proponent was reinforced by
the seemingly unfulfilled promise made therein that his “forthcoming book will
attempt to make explicit the principles of the implication-realisation model”. Now,
however, there is no excuse for such a view. The Analysis and Cognition of Basic
Melodic Structures (1989) established a comprehensive foundation for the detail of
his theories; now, The Analysis and Cognition of Melodic Complexity builds on its
predecessor and bids fair to present one of the most complete theories of music
cognition of recent decades.

It is not possible to review Melodic Complexity (hereafter ACMC) in isolation from its
predecessor, Basic Melodic Structure (hereafter ACBMS). The themes of the two
volumes are intimately interwoven, and the material presented in ACMC is heavily
dependent on ideas introduced in ACBMS. Indeed, although ACMC starts by
summarising those aspects of the theory introduced in ACBMS , I suspect that the
condensed nature of the presentation here would be of more benefit as reference
material to one who has already read ACBMS than to someone encountering the
theory for the first time. A brief account of the fundamental tenets of the theory as
presented in ACBMS will be provided before the enriched theory as expounded in
ACMC is discussed.

Narmour’s theory treats melody primarily as a note-to-note phenomenon, as a

“continuity of melodic relations whose intelligibility fundamentally derives from
lower-level bottom-up structures”. This, he feels, is justified because (ACBMS, p331)
“it is the dynamic non-reductive individuated melodic motion on the note-to-note
level that captivates music lovers.” In effect, the analytic method that Narmour
expounds in these volumes is intended to track and to reflect the fluctuations in
arousal that arise from the validation or invalidation of the moment-to-moment
subconscious inferences made by a listener concerning the way that a piece of music
will unfold.

Narmour starts by defining the bases for differentiating between simple melodic
patterns in terms of their implications for continuation. His theory is predicated on
a small number of general “hypothetical constants”. Two formal hypotheses - that
similarity between elements or events implies continuing similarity (A+A implying
A), and differentiation implies continuing differentiation (A+B implying C) -
together with concepts of “closure” and of “syntactic parametric scale” produce five
kinds of “melodic archetype”, of which some can be combined to generate a further
five “archetypal derivatives”.

1
In differentiating-out these archetypes and archetypal derivatives, he applies his
hypotheses separately to intervallic motion, register and pitch specificity, using
pitch, interval and registral direction as separable parameters. Within his theory, the
implication of continuation (“process”) arises through the operation of Gestalt
principles of similarity, proximity or common direction when consecutive pitches lie
in relatively close proximity to one another. Proximate pitches imply continuation
of pitch direction (“register”) and size of interval; thus the implication of an
ascending major second is of continued ascent by a small interval. The implication
of differentiation (“reversal”), arising when consecutive pitches span a large interval,
is conceptually symmetrical to continuation in that the registral and intervallic
implications and resultant syntactic function of reversal are opposite to those of
continuation. Two pitches spanning a large interval will thus imply reversal of
direction of pitch movement and differentiation in size of interval, yielding, as the
implication of (e.g.) an ascending major sixth, a descent by a small interval such as a
major second.

As degree of similarity (A+A) or differentiation (A+B) varies in melody between

patterns and elements and degree of implication fluctuates, he hypothesises a
syntactic parametric scale that is (ACBMS, p4) “an automatic input system that
determines what is similar or different, closural or non-closural function…and the
extent to which a melodic pattern is closed or open”. He suggests that “any pair of
melodic pitches transmits separate intervallic and registral messages to the
listener…a small melodic interval generates registral and intervallic implications of
similarity , while a large interval generates intervallic and registral implications of
differentiation or reversal”.

He puts forward a “syntactic parametric scalar” ordering of intervals from small to

large, implication of continuation gradually yielding to implication of reversal as
pitches making up intervals became less proximate and more differentiated. If
intervals differ by a minor 3rd or less than they are deemed “similar”, while if they
differ by a major 3rd or greater then they are “differentiated”. The size of the
difference thus determines whether a pattern is one of similarity or differentiation,
although different criteria for intervallic differentiation are adopted when registral
direction changes; when registral direction is reversed, a difference of ±major 2nd
or less is now regarded as a “similar” interval, whereas a difference of ±minor 3rd
or more is now “differentiated”. He suggests that this is because registral change
itself adds differentiation to the intervallic pattern. An analogous syntactic
parametric scale is proposed for register, going from lateral to lateral (sameness),
ascent to ascent or descent to descent (similarity) to ascent to descent, ascent to
lateral, etc. (differentiation). The reason for having separate parametric scales
governing interval and register is to enable the quantification of melodic motion, so
that while, say, a succession of ascending minor 2nds is going nowhere intervalically
(i.e., no differentiation is occurring between successive intervals) it is going
somewhere in terms of register.

Narmour’s five kinds of melodic archetype are: (i) process or iteration (symbolised
respectively as P and D);(ii) reversal (symbolised as R); (iii) registral return (aba);
(iv) dyad (two-element groupings, the unrealised implications of i and ii); and (v)
monad (one-element groupings, closed or unclosed with no generation of
implication). The five archetypal derivatives are based on process and reversal, each
constituting a partial denial of the realisations implied by the initial interval. These
are: (a) intervallic process (symbolised IP, a small interval to a similar small interval
in a different registral direction); (b) registral process (VP - small interval to large

2
interval, same registral direction); (c) intervallic duplication (ID - small interval to
same small interval, different registral direction); (d) intervallic reversal (IR - large
interval to small interval, same registral direction); and (e) registral reversal (VR -
large interval to even larger interval, different registral direction). The symbol “V”
here stands for register.

Narmour integrates the idea of “closure” into his discussion of these theoretical
archetypal patterns, process being non-closural and reversal being fundamentally
closural. Closure functions as the mechanism whereby multiple levels arise in the
analysis and cognition of melody. In his theory, “closure” is defined in terms of the
termination, blunting, inhibition or weakening of melodic implication, brought
about by its realisation or its denial. Strong closure produces the “transformation”
of notes (that is, their selection as events functioning at a structural level above that
of the musical surface), moderate closure brings about “formation” - “closure that
bids fair to generate a new level but does not actually create one” - and weak
closure simply gives rise to various degrees of melodic articulation. Closure here
largely determines the way in which patterns of melodic notes may be grouped, but
Narmour’s theory is more than a theory of grouping; as he suggests (ACBMS,
p104) “the nature of the [melodic] “trip” is as important as how it begins and ends”.

This caveat notwithstanding, closure here generates higher levels of structure, the
level immediately above that of the musical surface typically consisting of the notes
that begin and end each pattern in that surface. An example should help clarify the
operations involved. An isochronous sequence of notes such as b'-d"-f#"-g"-f#"-c#"-
f#" could be parsed as Process (P, b'-d"-f#"), Intervallic Duplication (ID, f#"-g"-f#"),
and again Intervallic Duplication (ID, f#"-c#"-f#"). Note that each successive pattern
starts with the pitch that ended the previous one. The “transformed” pitches - i.e.,
those deemed to function at the immediately superior structural level - would be the
first and last pitches of each successive, and hence closed, group, giving the series of
notes b'-f#"-f#"-f#" as the sequence’s representation at the structural level
immediately above that of the musical surface. The structural description of the
higher-level sequence itself might be a Reversal (R, b'-f#"-f#") combined with (as
opposed to succeeded by) a Duplication (D, f#"-f#"-f#"), by virtue of sharing an
interval (unison). Narmour makes use of the idea of combining in the analyses he
presents in ACBMS, but only explicitly lay out the conditions under which it occurs
in ACMC. Hence, the principles by which patterns can be parsed may be applied
recursively, each application generating a higher-level description.

Narmour makes clear that closure is not all-or-nothing; different degrees of closure
are exhibited by each of the archetypal patterns. As he points out, (ACBMS, p337)
“degree of closure is dependent on functional differences between first and second
interval. The greater the amount of initial implication, the more closure occurs. the
greater the amount of implication in the terminal interval the less closure occurs.”
He provides (ibid., p361) a table of archetypal patterns ranked from most open to
most closed, in the order VP, P, D, VR, IP, ID, IR, R, ranked in terms of their
intervallic motion and their registral motion.

Implication functions here not only prospectively but also retrospectively.

Prospective interpretations of implications arise when the implied realisation
actually occurs, and retrospective interpretation of implications when the implied
realisation does not occur. An instance of this would be when an interval that
normally implies continuation realises a reversal pattern. Thus the difference
between an IP - small interval to small interval, different direction - and (R)

3
(retrospective reversal, the retrospection symbolised by the parentheses) - small
interval to small interval, different direction - is that IP occurs when intervals are
close together in size (i.e. similar) whereas (R) will have occurred when the intervals
are sufficiently differentiated as in, say, a leap upwards of a major third followed by
a descent of a minor second.

Apart from intervallic and registral motion, a number of other factors interact in
determining how a melody may be parsed into contiguous archetypal patterns.
Metre and duration play crucial roles. As he says, “Metric accent always has an
articulating effect on melody…metre also has the potential to combine with other
parameters to cause the transformation of melodic tones to higher levels”. For
metre to play such a role it is required that its establishment be unequivocal.
“Durational cumulation” (i.e., a long note succeeding a shorter one, relative
durations standing in - at least - the ratio 1:2) can convert the last note of the final
interval of a pattern from an open pitch into a closed one. Metre figures in the
closural evaluation of durational cumulation in that (i) it indicates to the listener
whether the metrically specified place of the realisation has passed by or not, and (ii)
metric stress figures in the perceptual evaluation of cumulation in the sense of
emphasising the long note and thus marking it for closure on the level of its
differentiation. Dissonance may also act alone or interact with cumulation in terms
of determining degree of closure or continuation.

Narmour differentiates sharply between what he sees as the immutable properties

of the basic “archetypal patterns” and any specific stylistic implications that these
might be identified as having. The functioning of the archetypal patterns is referred
to “reflexive” bottom-up processes, whereas any invocation of stylistic
considerations is held to be “top-down”. In his words (ACBMS, p9), “In the
implication-realisation model we keep the top-down mapping of style separate from
the bottom-up input of the primitive shapes making up the style…from the
listener's perspective, style is thus replication of experience, instantiated learning
that leads to inductive inference...It operates directly on learned structures of
relations influencing input shapes like single melodic intervals only as such shapes
occur within learned style structures.” For Narmour, “style shapes” are the
properties that individual parameters exhibit, while “style structures” are complex
patterns organised by many different parameters.

The efficacy of the primitives of the theory - implications of similarity,

differentiation, and parametric scales - is referred by Narmour to their instantiation
in bottom-up processing: in his words, “a kind of automatic cognizing…hard-wired,
innate and unavoidable in its consequences” wherein “the listener subconsciously
constructs her musical perception solely out of the individual parametric features
presented to her”. However, he warns (ACBMS, p88) that the operation of the
primitives of the theory “must not be invoked simplistically because style structures
may override style shape under the right circumstances”. In other words, despite
the specificity of the hypotheses that he outlines in his theory, he leaves considerable
scope for interpretation of implication on the basis of style, whether intra- or
extraopus (the influence of both levels of style being symbolised within his theory
as, respectively, “os” and “xs”). In fact, most of the analyses that are presented in
the course of the two volumes are employed by him to demonstrate the potentially
determinant role that style may play in shaping melodic implication. It should be
noted in this context that his theory treats the “scale-step” properties of pitch as
being stylistically derived. He states (ibid., p82) that “In terms of traditional tonal
music theory my argument is that the stylistic scale step functions usually assumed

4
about tonal music operate as a learned top-down schema influencing the intervallic
implications hypothesised in the parametric scale of the model”, and proposes that
scale-steps are nothing more than top-down atomic schemata (ibid, p85) and “thus
no more deserving of preferential treatment within the implicative theory than any
other form of learning”.

ACBMS has largely dealt with successions of archetypal shapes at the level of the
musical surface - “consecutive structures” - outlining the structures that may occur
and the ways in which they succeed one another. ACMC is primarily devoted to the
exposition of more complex relations than simple succession between temporally
contiguous patterns, as well as to the way in which these successions of temporally
contiguous patterns give rise to higher levels of structure, or, to use his term,
“transformational” levels. The two aspects of his theory that form the subject-
matter of the book are intertwined in that the principles that govern the ways in
which patterns succeed (or link with) one another at the musical surface also
determine the higher-level “transformed” structures that may emerge, though it
must be noted that in his theory, “the word structure is not synonymous with
transformation to a new level. Structuring in a complex syntactic system like music
exists at all levels.”

Narmour in fact introduces some aspects of “transformed” - i.e., higher-level -

structuring in ACBMS. He proposes a set of “rules of production” (ACBMS, p412),
which state that “when closed, the initial and terminal tones of processes,
duplications and reversals of all kinds (P, IP, VP, D, ID, R, IR, VR) lead to dyads that
may imply and realise continuations and reversals on higher levels. On the other
hand, the transformed tone of any dyad always leads to a monad which may
connect with other structural tones to create implications and realisation of all
types.” Despite these ostensibly simple production rules, the musical surface does
not constitute a “diminution” (to use the word in its Schenkerian sense) of higher-
level structure in his theory; he is emphatic on this point, asserting (ACMC, p255)
that “Surely, the point of tonal music is not just to prolong a few idealised, top-
down, consonant chords”. He further clarifies this in stating (ibid., p352) that
“Because musical levels are hierarchical rather than systemic, the higher-level pitch
patterns that arise from lower levels embody rather than assimilate implications,
realisations and denials at lower levels.” In this, he is again asserting the primacy
and autonomy of the bottom-up processes operating at the level of the musical
surface.

Narmour states (ACMC, p46) that his “main goal in this volume is to explore how
the primary and derivative archetypal structures combine into various two-
structure groupings and to explain how these in turn create longer structural
chains.” He stresses that non-closural contexts cause combining and chaining of
structures. These structures - the archetypal shapes or archetypal derivatives -
succeed one another by sharing a pitch, the terminal pitch of one being the initial
pitch of the next (as outlined in ACBMS). Two such structures - e.g., P and IP - may
combine if the terminal interval of one is the initial interval of the next, or more
than two may chain by sharing linking intervals, the conditions that determine
combining and chaining being largely metrical/temporal “low-level conditions”,
such as the occurrence of a dissonance on a metric accent, the presence of ongoing
meter etc. Dyads and monads never function as links in a chain, as individual
structures in a chain must have at least three notes.

5
He outlines, on note-to-note level, the eight low-level conditions that create
combining and chaining:
1) melodic dissonance on metric accent
2) lack of metric accent
3) processive envelopment of metric accent
4) duplicative envelopment of metric accent
5) harmonic processive envelopment of metric accent
6) durational processive envelopment of metric accent
7) dynamic processive envelopment of metric accent
8) ongoing syncopation in isochrony
Conditions 1) and 2) are “the deformed oppositions of harmonic, metric and
durational influence” (the other-parametric conditions that would otherwise cause
closure, here working to create or reinforce conditions of non-closure), whereas 3)
to 8) are, in effect, the realisation of implicative continuation. In all these conditions
it is non-closure that causes discrete structures to combine and chain together.

In the case of succession of discrete melodic structures, the initial and terminal notes
of each structure (i.e., those shared by consecutive structures) will function as
“transformed tones” on a higher hierarchical level. In combinational patterns and
chains, however, the initial and terminal tones of the overall pattern will function as
transformed tones, “non-closure enveloping all tones between these two anchor
points”. Thus a sequence formed by a succession of patterns is likely to give rise to
a higher-level representation containing more “transformed” notes than would a
sequence comprising combinations or chains of patterns. However, it should be
noted that although in chains fewer tones will be “transformed” to a higher level,
structure within the chain will be delineated by medial tones exhibiting greater or
lesser degrees of articulation (being unequivocally located at the lower level) or
formation (formational tones being those that portend a higher hierarchical level
but nevertheless remain at their level of occurrence).

Narmour starts his account of combined patterns by introducing “Combinational

structures of contrasting parentage”, covering combinations of process followed by
reversal (PR, PVR, DVR, PIR, IPR, IPVR, IPIR, IDR, IDVR, IDIR, VPR, VPVR and
VPIR) wherein each element may be in either prospective or retrospective form.
Here, he outlines the conditions that bring about the theoretically simpler forms
involving combinations in which closural “reversal” structures succeed and combine
with (by sharing an interval) non-closural processive or duplicative structures. In
the course of this, he points out that patterns such as DR and DIR are theoretically
irrational and thus analytically impossible. This is because the D structure
(Duplication) involves note repetition (and hence repetition of the interval of a
unison), while R and IR structures both require a differentiated decrease in interval
size from first to second interval, not possible if their first interval is a unison. The
issue of whether two pattern succeed one another or combine is dependent on
whether his eight conditions (above) can be deemed operational or not. The issue of
whether any combination is between patterns in their prospective or retrospective
forms - e.g., as PR or P(R) - is largely dependent on the degree to which each of the
constituent patterns was anticipated or was unexpected. While the nature of the
basic patterns involved may play a role in this, it is often the influence of style -
particularly intraopus style (symbolised by him as “os”) - that is the determinant
factor. Thus (ACMC, p52) “Process/reversal combinations at beginnings of pieces
tend to be P(R) [i.e., the reversal is unexpected and thus retrospectively identified as
a reversal], while subsequent ones become PR with (os) attached, signifying the

6
influence that intraopus style has on our changing assessment of retrospection and
prospection.”

In deciding how patterns are linked, it is important to determine the level at which
each pattern is functioning. Two patterns functioning at the same level of
implication (typically, metrically-defined) may succeed each other or combine, but if
the second of two apparently successive patterns is at a metric or durational level
below that at which the principal events are being articulated, it may well be
“embedded” within a superordinate structure rather than combined with the
preceding pattern. As he states (ACMC, p103) “Combinational structures often have
lower level patterns embedded within them. Embedding comes about when
realisation of lower-level patterns takes place before the implied realisation of
higher-level patterns. Such two-tiered patterning is possible because melodic
implications specify not only registral direction, intervallic size and specific pitch
(when mode is known) but also durational length and metric location of realisation.”

Having dealt with combinational structures formed by ordered pairs of open and
closed patterns, Narmour then accounts for the conditions that can give rise to a
partially closed reversal pattern combining with a subsequent non-closural one, as in
an RP combination. He suggests that for this to happen the closural note of the
reversal must be weakened by the occurrence of one or more of his eight
“combining” conditions, and that these same conditions must also suppress the
primacy of the initial tone of the process. If such conditions pertain then the
terminal interval of the reversal will “dovetail” to the initial interval of the process.
He has, in fact, referred to such combinations in accounting for “gap-fill” melodies in
ACBMS (p220, et seq.). An exposition of “same-parentage” combinations then
follows, describing the conditions that underlie the combining of two consecutive
non-closural processive or duplicative patterns (P, D, ID, IP and VP) as well as of two
closural reversal patterns (R, IR, VR). Again, these patterns may combine in
prospective or retrospective interpretations, factors governing the interpretation
deriving either from the specific content of the patterns or from stylistic influences.
His account of the forming of combinations is summarised in a figure (ibid., Example
7.13) illustrating the principles that govern the interaction of meter and dissonance
in the making of either separate structures or combinations.

Section 3 of ACMC provides a detailed account of chains, which are formed when
three or more structures interlink by sharing intervals. While archetypal structures
and combinational derivations are finite in number, totalling 235 (see his Appendix
5), in melodic chains the possible concatenated orderings of continued interval
sharing are limitless. He states (ACMC, p138) that “Chained melodic structures
result when dissonance, durational patterning (frequently additive) and metric
emphasis are sufficiently non-congruent with one another to dovetail several
different patterns.” The eight conditions that underlie combinations (see above) also
cause chaining, which results from continuing non-closure preventing hierarchical
transformation of the initial and terminal pitches of the constituent patterns.

As in his discussion of basic patterns and combinations, Narmour warns that chains
cannot be identified by “simplistic” search procedures conducted at each metrical or
durational level. Dissonances may displace the implied location of a melodic
realisation, and hence “chains frequently mix different durational levels.” As he says
(ACMC, p186), although “the general rule that ongoing metre in additive durational
[isochronous] patterns causes combining and chaining holds…it is subject to
numerous contextual qualifications. One cannot apply it - or any other rule of the

7
implication-realisation model - mechanistically. Indeed, the bottom-up/top-down
orientation of the theory frames cognitive reality hypothetically rather than
axiomatically, insisting on the cognitive analytical probability of rule violation.”

Chains are “runs” of interlinked (i.e., interval-sharing) patterns at the same

hierarchical level. The incorporation of the concepts of combinations and chains in
his theory allows for the description of complex melodic structure at the level of the
musical surface, providing a means of delineating the internal structure of
potentially long sequences of events lying between the two “transformed”
(structural) tones that bound a basic pattern, combination, or chain. Combinations
and chains permit a coherent account of the microstructure of melody in terms of
articulations and formations (portents of closure that do not give rise to higher-level
transformations) to be provided by the theory. Common music-theoretic entities
such as the appoggiatura are thus explicable as functional elements within chains;
“the appoggiatura is a chief cause of structural chaining in melody when resolution
of the dissonance is metrically and harmonically weak” (ACMC, p190), the
appoggiatural note itself frequently functioning formationally.

Narmour stresses the importance of “other-parametric” interference and interaction

in creating chains, indicating that (ACMC, p207) “In chains, the non-closure of the
harmony (an implied or realised dissonance) versus the closure of metre (an accent)
and the closure of harmony (a weak resolution) versus the nonclosure of metre
(ongoing non-accents) offset one another. It is such offsetting conditions that keep
the pattern ongoing.” He makes explicit the ways in which the manifest structural
complexity that chains display in his theory is primarily a characteristic of the
musical surface, stating that (ibid., p199) “In moving hierarchically from lower levels
to higher levels, one will, of course, observe complexity giving way to simplicity as
lower level articulative and formational functions of tones are subsumed within
higher level transformational [structural] functions.”

Narmour is concerned to demonstrate the potential utility of the theory.

Throughout its exposition, he is prepared to draw inferences not only about the
perceived structure of melody but also about compositional usage and intent in
employing particular structures. For instance, he suggests (ACMC, p177) that the
“most common chain is that in which a process functions medially, flanked on both
sides by contrasting structures”, and that (ibid., p193) “one common syntactic
strategy is to vary structures of the chain from partial intervallic realisations of
implication at the beginning, to full realisations of implication in the middle to partial
realisations of implication that intervalically increase at the very end.” Similarly, he
proposes that the typologies available within his theory can help to clarify the details
of the process of historical style change in music. He claims (ibid., p254) that because
“in a general sense I conceive of articulation, formation and transformation as a
spectrum of bottom-up relations across which I imagine a flexible ordering and
ranking of the hierarchical functions of tone”, it is thus possible to (ibid., p185) “chart
the structural evolution through melodic tones of various strengths…tracing genetic
order from articulation to formation to transformation to deformation.”, showing
how “the syntactic structures of music gradually undergo historical change”.

Given that combining and chaining are largely motivated by metrical and harmonic
factors Narmour devotes space to detailing the dynamics of this interaction, putting
forward classifications of dissonance and resolution and outlining how harmonic
factors can, either in conjunction with metre or separately, act to delineate melodic
structure (see also Appendix 2, p374). He posits three classes of dissonance, strong,

8
medium and weak, which he symbolises respectively as _, x and (x). He suggests
(ACMC, p210) that “This threefold division is a bottom-up hypothesis not specifically
tied to any restricted conception of tonal style…all that matters is that the invoked
style allows for (1) a prospectively perceived vertical (acoustical) difference between
consonance and dissonance (2) a perceived difference between chordal and non-
chordal properties and (3) a perception that a particular group of tones is a set
distinguished from other collections of tones and thus other modes.” An instance of
his strong dissonance is one that is a non-chordal dissonance lying out of the key
and occurring with a chord change at the moment of the dissonance itself; this he
terms “contramodal”, to indicate that it runs counter to the established mode.
Moderate classes of dissonance include diatonic non-chordal appoggiaturas over a
change of chord introduced melodically by a small interval, and non-chordal
diatonic suspensions with a change of chord at the tie. Weak dissonances are of two
types (ibid., p212): “first is non-chordal dissonance occurring over the same chord
(passing, neighbour, escape tones): the second covers all essential chord-tone
dissonances with or without a change of chord - added sixths, sevenths, ninths, even
elevenths and thirteenths in the right circumstances regardless of whether such
tones are diatonic or contramodal”. His definitions of dissonance are explicitly
independent of metric position, being reliant on chord change which he states
creates “metric emphasis”. Despite the specificity of the definitions, he does warn
(ibid., p214) that “The description of weak, moderate and strong dissonance is
variable”.

Balancing his three types of dissonance we find three types of resolution, again
weak, moderate and strong. In a strong resolution, the chord change occurs
precisely at the consonance, in a weak resolution no chord change takes place on the
consonance, while a moderate resolution lies between these extremes. From this
hypothesis of three types of resolution and dissonance, nine possible dissonance-
resolution pairs can be formed. He further characterises these types of
progressions, stating (ACMC, p216) that “Weak dissonances or resolutions are those
that occur with little other-parametric differentiation, i.e. little metric or textural
stress. Strong dissonances possess much differentiation, and high differentiation
makes for strong resolution as well. Moderate dissonances are ones that do not
belong to the chord but at least belong to the key. Moderate resolutions move with
the change of chord but in a rather weak progression and ones involving at least
two common tones.” Despite the pervasive influence of harmonic structure and
processes in constituting chaining, harmonic influence is not in itself a necessary
factor in chaining. As he says (ibid., p226) “[although] normally, chaining occurs
because strong dissonance and weak resolution deform the metric organisation
…chaining can take place in the absence of dissonance if the melodic patterning itself
deforms the metre making the location of metric accent ambiguous.”

From p219 to p221 he provides the most extended analysis presented in either
volume, contrasting Salzer’s (1962) “Schenkerian” analysis of the first four bars of
Mozart’s Sonata K311, 2nd movement, with his own implication-realisational
analysis. He applies his theory to each voice independently and separately,
demonstrating how his theory can account for both the musical surface and higher
structural levels in terms of melodic implication. He criticises Salzer’s analysis for
“showing the phrase as one more instantiation and confirmation of tonality”,
contrasting it with his own which he suggests “shows it as an example where
melodic, metric and durational weakening of prolongation of the tonic occurs.” He
claims (p221) that while “Schenkerian (Salzerian) theory aims to reduce foreground
levels to past learning, to previously known style structures, the implication-

9
realisation model insists on the possibility that bottom-up low-level patterning
permanently has the power to deform these prior invocations”.

The penultimate section of this volume is devoted to the issue of “structural

networking”. He deals first with co-existent but mutually irreducible types of
structure, which may arise when theoretical primitives that are independent within
the theory, yet share fundamental characteristics, can be applied to the same musical
material to generate alternative but co-existent representations. This is one of the
ways in which his theory appears to deal with structural ambiguity; these
“networked structures” do not play a role in the transformation of tones to a higher
level, but exist concurrently with the structures and procedures that do fulfil this
role, giving rise to “tangled hierarchies”. As he puts it (ACMC, p232) “The
implication-realisation model everywhere insists on non-congruent networked
possibilities”. For example, one of the archetypal patterns - aba, or registral return -
is always incorporated in the archetypal derivatives IP and ID, but can occur by itself
without requiring these patterns, often in overlap with other, more structurally
significant, patterns. He notes (ibid., p231) that “nowhere is this independence [of
aba] more felt than in complex harmonic contexts where its termination as a
dissonance lies noncongruent with the metre.”

He then proceeds to give a detailed account of the conditions under which tones that
are dissonant at the musical surface may nevertheless function at higher levels -
“transformational dissonances”. In this, the theory presents a view that is at odds
with much of “conventional” music theory, as he notes. He puts forward a scheme
for describing the interaction between temporal (i.e., long note) accent and
dissonance under a variety of conditions of non-congruence between parameters
(e.g., when a strong and hence non-closural dissonance occurs on what would, by
virtue of its length in context, be a closural duration). The system that he proposes
(ACMC, p249) outlines the circumstances in which dissonant tones should be
construed as either articulations, formations, or transformations (all other things
being equal). He suggests (ibid., p248) that “The stronger the dissonance, the more
cumulative must be the duration [i.e. the greater must be the positive difference
between the duration of the note on which the dissonance occurs and that of the
note immediately preceding it] in order to effect higher-level transformation”. In
the context of these proposals he is able to state (ibid., p251) that “appoggiaturas set
in additive [isochronous] durations are wholly-nonstructural; those set cumulatively
are either articulative, formational or transformational, depending on the degree of
dissonance and the degree of cumulation.” and to claim that “The top-down stylistic
view that lumps all these into one class ignores how the bottom-up parametric
context structurally differentiates them”.

A further chapter explores other potential “networked structures”, dealing with the
“post-denial” processive implications of tones, those implications that persist
beyond their denial to the extent that they create alternative structures (generally at
the level of the musical surface) that are co-existent with those that give rise to
transformations. He holds that it is necessary to recognise the existence of such
discontiguous realisations (ACMC, p260) “because processive melodic implications
remain alive during the very moment of their negation”.” These “post-denial”
implications he refers to as “time-tagged”, meaning by “time-tag” the duration by
which a melodic implication persists beyond its denial. He suggests that, all other
parametric things being equal, the time tag of a process lasts approximately two to
three times as long as the last tone of the interrupted implicative pattern, (the
further away being the discontiguous connection, the weaker the connection), and

10
states (ibid., p264) that such time-tagged structures are “always simultaneously
secondary structures, functioning in subservience to the primary structures
constructed of contiguous connections”.

The concluding section of the volume starts with a chapter considering the ways in
which his theory deals with melodic grouping. As he says (ibid., p293) “All analytical
theories entail concepts of grouping. A rule that prescribes the conditions under
which a given analytical symbol operates in a syntactic system is essentially nothing
more than a rule of grouping.” The most important factor in determining grouping
in his theory is closure. In general, within the implication-realisation model, a tone
that ends a group (basic pattern, combination or chain) on one level will initiate the
subsequent group. Melodic closure (which may be determined by interaction of
parameters other than pitch and register, such as duration, metre or harmony) will
result in what he terms a perceptual and analytical “startover” on a tone of closure.
However, two consecutive groups may be separate in not sharing a common tone
as end and beginning; “If a tone of closure and the next following tone are not level
equivalent, creating the denial of an implication (and no pattern of registral return,
aba, is present) then listeners will tend to forget the unclosed dyadic relation and
separation between the two tones will occur” (ibid., p298). Such a separation
between groups may also arise as an effect of “modelled repetition” (described
earlier on pp129-132), wherein the influence of intraopus (“os”) or extraopus (“xs”)
style enables a listener to anticipate the immediate repetition of a sequence already
identified as a group.

Narmour points out (ibid., p312) that “What makes the analysis of grouping such a
difficult theoretical subject is that music displays both continuous and discrete
properties. Ongoingness simultaneously exists with formal disjunction.” The
capacity of his theory to cope with both ongoingness and disjunction - though
focusing largely on the representation of ongoingness - means that grouping in the
theory can take many forms, and that the theory is flexible enough to accommodate
a wide range of musical textures and articulations of structure. He suggests (ibid.,
p312) that while “the norm is for low-level patterns to emerge as separate formal
entities while high-level relations establish uninterrupted continuity” it is possible
for “high-level events to exhibit much discontinuity while low-level ones remain full
of connections”, both extremes being representable within the implication-
realisation model. As he says (p352) “melodic analysis and formal analysis
are…indissolubly linked. One cannot properly explain higher-level melodic
structures without taking into account their lower-level forms.”

His concluding chapter is largely concerned with establishing the utility of the
theory and delineating the logistics of its application (including suggestions for some
possible variants of the theory). Starting by outlining various formal symmetries
that exist between components of the theory, he then seeks to make explicit the
value of the theory considered from a music-theoretic and musicological
perspective. He suggests (ibid., p336) that “One value of the symbols advocated
herein is that they enable us to discover and identify the common ‘word orders’ of
melodic ‘sentences’…we have already seen a number of common structural
orderings in the ‘sentences’ making up melodic chains, and have shown how such
sentences characterise the compositional strategies found in melody…A string of
structural symbols therefore represents a syntax, ordered tokens typifying the
melodic language of a given composer, a given work, a given genre or any other
level of the style hierarchy.”

11
He also refers to the possibility of using his system on computer because of the
“economy” of the symbols, essentially as an element of a database-type application,
handling - and enabling the cross-comparison of analyses of - large amounts of
musical data. In this respect he suggests a number of extensions to the symbology
of his theory in order to represent, e.g., registral direction (which is not explicitly
encoded in the basic symbols), and the separation or conjunction of basic patterns.

He then examines what he conceives to be the limitations, strengths and possibilities

of the theory, pointing out (ibid., p346) that while “The analytical theory offers a
way to study the multi-faceted replacement techniques of melodic variation and the
kind of structural changes that varied repetition brings about”, nevertheless “the
employment of analytical strings of note-to-note relations may lead to erroneous
and misleading comparisons unless we accurately track levels and keep them
hierarchically separate.” One of the principal benefits of his theory is, he feels, the
fact that it can model both the ongoing and the hierarchical aspects of music in
cognition; “Music is not either top-down or bottom-up, but simultaneously both,
lower levels influencing higher ones and vice versa…as music moves horizontally
through time it is simultaneously moving vertically, during which hierarchical levels
gather strength Thus perceptual ‘feedforward’ between low and high level and
cognitive ‘feedback’ between high and low level constantly takes place. Our
analyses and explanations should reflect this psychological state of affairs” (ibid.,
p349). He concludes by speculating on the appropriateness of his theory for
representing how listeners might encode melody in terms of functional similarity
and differentiation, raising questions - such as the influence of form on the analysis
of melody - that the theory might be developed to address, in the context of
outlining aspects of the theory to be covered in future volumes. A series of
Appendices presents summaries of his rule-systems and symbologies, a catalogue of
archetypes and combinational structures, and sample analyses of early Western
music and music from other cultures, which he deems his theory capable of
analysing because of its “general panstylistic outlook”.

On first acquaintance, this volume, like its predecessor, appears formidably difficult.
It presents a theory that is neither simply conceived nor simply expressed.
Nevertheless, even from the inadequate summary presented here it should be
evident that this theory could be invaluable in its precise depiction of the
ongoingness of musical perception; it attempts to encapsulate the continual flux
experienced by a listener as a piece of music unfolds and as perception unfolds with
it, sometimes drifting ahead of, and sometimes racing to catch up with, what has
happened, what is happening and what might happen. In the level of detail that it
achieves in accomplishing this and in its flexibility, the theory appears virtually
unique.

Having said this, the theory, and indeed, its presentation in this volume and its
predecessor, have several problematic aspects. While the matter of presentation
may seem trivial in respect of such a comprehensive and incisively conceived
theory, it is important; Narmour’s complex ideas warrant, indeed, require, a wide
dissemination in order that their import can be assessed in the different intellectual
communities whose central concerns they address. It seems to me that the principal
problem with this volume is that for a book introducing a new theory, it reads at
times as a slightly awkward amalgamation of textbook and justificatory manifesto:
textbook in outlining the elements and their interrelations in the theory and
manifesto in its provision of a dense analytical and cognitive contextualisation of
those elements and relations. This conflation of functions does not make the book

12
easy to approach; reading it, one experiences continual conflict between being
informed by the exposition of the theory and challenged by the invocation of a vast
range of music-theoretic and cognitive phenomena in its justification.

As an instance of this, while the detailed music-analytic examples are welcome, and
demonstrate the wide applicability of the theory they can interfere with the clear
presentation of the theory by continually making evident its “except-cases”,
loosening the reader’s grasp on the import of the theory’s fundamental principles
and premises. It might be argued that, after all, unless this theory can be applied to
real music, its validity must be questionable; on the other hand, the constant
validation and re-validation of the theory in terms of the analytical examples
presented seems paradoxically less convincing than the author might have wished,
perhaps because of the particularity - the idiosyncrasies - of each musical example,
which, while emphasising the theory’s wide musical applicability, act continually to
undermine the reader’s understanding of the general principles lying behind the
basic elements of the theory.

The issue of general style of presentation aside, there seem to me to be a number of

respects in which the theory is itself problematic; these may derive from the way in
which certain issues are presented, but may also reflect unclarities or contradictions
in the role and function of specific topics within the theory. The main difficulties
appear to lie in assessing the status of rhythm and metre, and of harmony, within
the theory.

While a considerable part of both volumes is devoted to specifying the operational

characteristics of metre and rhythm in the implication-realisation model, on the
whole the influences of metre and rhythm are neither clearly nor separately nor
separably delineated from the factors that he purports to demonstrate govern note-
to-note succession. Given the formality of much of the rest of the theory, his
treatment of rhythm and metre appears too discursive. In fact, his emphasis on
note-to-note relations (in terms of durational “cumulation” and
“countercumulation” relations) might be one of the factors that obfuscates the role
of duration in musical patterning; the potential for durational patterning and
metrical (accentual) organisation to act independently in structuring the progression
of musical time appears unexplored and largely unacknowledged here. At the least,
the admixture of rhythm and metre within the theory in influencing patterning
tends to undermine the potential “primitive” power (see, e.g., Sloboda and Parker,
1985; Jones and Yee, 1993) of metre (once established through intraopus or
extraopus style) as a structuring agent. It could have been helpful to articulate
theories of metre and rhythm independently before instancing the influence of
metrical and durational factors in the overall implication-realisation model.

Similarly, despite extensive coverage in the text, the role and status of harmony in
the theory remains ambiguous. What might be regarded as inseparable correlates
of harmonic tonality appear, at times, to be construed as functioning from the top-
down - as elements of stylistic and thus variable influence - while at other times their
influence is overtly bottom-up and “reflexive”. Thus Narmour’s designation of
dissonance as an operational bottom-up factor in forming combinations and chains
appears at odds with his identification of the “scale-step” functions of notes within a
mode as “top-down atomic schemata…and thus no more deserving of preferential
treatment within the implicative theory than any other form of learning”.

13
To be sure, it could be argued that this ambiguity is in fact resolved in that he lays
down ostensibly “style-free” conditions for the existence of dissonance within the
theory. In ACMC, in defining his three classes of dissonance (see above), he
suggests that these constitute a “bottom-up hypothesis not specifically tied to any
restricted conception of tonal style”. The existence of dissonance within any musical
style or genre is referable to the potential for three perceptually operational factors:
(1) differentiability of consonances and dissonances on the basis of “vertical
(acoustical) difference”, (2) differentiation between “chordal and non-chordal
properties” and (3) discriminability of groups of tones. However, there are
problems with these specifications.

To start with, his point (1) concerning a “perceived vertical (acoustical) difference”
between consonance and dissonance would appear to imply an almost Direct Realist
view of perception, i.e., the existence of a direct correspondence between the nature
of the physical signal and its perceptual (musical) significance. But without clearly
articulating a theory of how such a correspondence might operate such a view is not
tenable. In any case, as Helmholtz suggested in the last century, it is more likely
that the ways in which our sensory systems make use of the information embodied
in a physical signal act to determine its musical functionality than that such
functionality is referable to the nature of the physical signal alone. Reference to or
inclusion of some such psychoacoustical considerations (as, e.g., outlined by
Parncutt, 1989) might well “save the appearance” of this aspect of Narmour’s
theory, but such reference is not made here.

In the second place, his point (3) is dependent on the distinguishability in perception
of “collections of tones, and thus…modes”. While it is feasible to argue that
discrimination per se is not dependent on prior identification, it does seem probable
that discrimination between “collections of tones” requires the existence of a
relatively rich representation in cognition of interrelations between such “sets”.
Given that scale-step function - intrinsically tied to mode - is located in Narmour’s
theory as being active in the stylistic (top-down) domain, it is difficult to see just how
the capacity to discriminate between modes can be simultaneously bottom-up. I
suspect that this ambiguity arises because of the way in which Narmour chooses to
maintain a strict separation between bottom-up (invariant) and top-down (variable)
factors within his theory, a point that requires further consideration in the context of
a critique of fundamental aspects of the theory.

The first such aspect to be considered is Narmour’s employment of “syntactic

parametric scales” within each separate musical parameter in the implication-
realisation model. It is not so much the separation of parameters within Narmour’s
model that is problematic - indeed, it is largely the separate treatment of pitch and
register that endows the theory with its power - as the fact that the syntactic scales
used in conjunction with each separate parameter are largely analogous in structure.
Although he refers to the possibility that such scales might differ in structure and
function between parameters (see, e.g., ACBMS, p311), on the whole the scales that
he uses to order harmonic and melodic (interval) movement are unidimensional (for
harmony, see ACMC, p127, fn 6). Whilst the adoption of such an approach could be
justified on theoretical grounds (“rationalistic theoretical construct”), it does appear
to underrate severely the degree of complexity that has been shown to be
operational in the adjudged perceptual similarity between pitches (see, e.g.,
Krumhansl, 1990). Narmour might choose to reject such a complex basis for his
syntactic scale of pitch on the grounds that it is largely a product of formal learning
(as Krumhansl herself seems to suggest) and hence “top-down”. However, such a

14
contention brings one to the heart of the problem that Narmour’s theory -
considered as a cognitive as opposed to analytic theory - presents; is his distinction
between “bottom-up” and “top-down” processes as rigid and unchangeable as he
presumes ?

If a rigid separation prevails, then the contradiction noted earlier, concerning the
simultaneous top-down and bottom-up operation of tonal harmonic factors remains
unresolved. If, on the other hand the separation between top-down and bottom-up
- between learned and hard-wired - is not so rigid as Narmour proposes, does his
theory necessarily fall ? I think not, despite the fact that one of his principal
concerns is to separate out the influence of variable stylistic factors from the
determining influence of general cognitive propensities. For one of the most
evident facts about neural systems on the whole is their plasticity, their capacity to
adapt and change as a function of learning, even at the level of “neural wiring” (see,
e.g., Rose, 1993). Thus, if a sufficiently common basis for learning could be
postulated, and it could be shown that such learning resulted in the general
operation of cognitive processes that were effectively reflexive and not amenable to
conscious introspection, the question of why such processes should not be
considered as effectively equivalent to those that are hard-wired and hence bottom-
up would have to be answered.

Why should the results of such common learning experiences be excluded from
playing a formative role within the bottom-up elements of the theory ? Narmour
does consider and reject this possibility, suggesting that “Although all experienced
listeners share style knowledge to some extent, it is not tenable to posit that any one
cultural segment of that shared knowledge represents a perceptual whole. Just as
there’s no such philosophical thing as a language, so there is no such cognitive thing
as a style.” In making this statement, Narmour does not appear to consider the
possibility of trans-genre stylistic constraints that may be orientated around some
constant structural core - largely, what the role of tonality can be conceived of as
being in the common-practice period, at least, as its historical traces may exist for
and in our perceptions now, in the present, through our exposure to the music of the
last five centuries.

Indeed, to accept such a possibility would severely limit the scope and applicability
of the basic principles of his theory. They could not, for instance, be applied
hermeneutically in the exegesis of compositional practice and intent. But it would
not disqualify the theory from elucidating the experience of melody in our time and
within what might be described as Western culture. In fact, the incorporation of
such considerations would vastly enhance the explanatory power of the theory, but
only in respect of this restricted domain. Such a common basis of learning - a
common realm of experience - is, in many ways, no mere hypothesis. Its operation
in respect of music can be and has been empirically demonstrated in a number of
domains (see, e.g., Sloboda, 1985; Trehub and Trainor, 1993), as has its generally
non-introspectable nature (see, e.g., Cross, Howell and West, 1983; Lamont and
Cross, 1994) - but only within a particular, though increasingly global, musical
culture. Indeed, attempts to predict and predicate the nature of music and musical
experience within other cultures on the existence of cognitive universals seem
doomed to failure; recent personal experience (Stobart & Cross, 1994; Cross and
Stobart, in preparation) has inclined me to believe that cultural particularities are the
primary determinants of musical expression, often subverting the operation of
cognitive universals in the cause of articulating a culturally acceptable and distinctive
aesthetic. While the operation of many universal cognitive principles may be

15
adduced or invoked in providing an explanation of the listening process, it cannot be
assumed that these cognitive principles will operate in the same way - i.e. will be
expressed in any way that we can currently empirically observe - within the confines
of a different culture (whether that difference is historical, geographical, or socio-
economic) unless those cognitive principles can be shown to be so deterministic that
their modes of operation are confined to an extremely narrow range. The only way
that the sort of approach to the exegesis of cultural artefacts that Narmour espouses
could work is in the context of a “deep” immersion in and study of the cultural
matrix of that artefact (cf Blacking, 1981).

In effect, the incorporation of certain trans-genre stylistic constants - such as some

aspects of generalised scale-step function, perhaps along the lines sketched out in
ACBMS, pp82-85 - would clarify the operation of Narmour’s theory though
restricting its scope, at least insofar as the basic theory, in terms of bottom-up
proclivities, has any hermeneutic power that can elucidate compositional intent. But
that hermeneutic power derives largely from the stylistic knowledge - the “degree
of cultural immersion” - that an analyst brings to an analysis to complement the use
of the theory, as Narmour himself indicates. As he says (ACMC, p59), “A given
analyst may assert that style impinges on, and thus changes the meaning of, an
implication normally governed differently by the if-then perceptual constants of the
theory. But when such assertions are not self-evident it is the responsibility of the
analyst to offer evidence demonstrating that this or that implicative pattern
conforms sufficiently to this or that stylistic continuation.”

This is a highly developed theory that is full of rich possibilities, as much for the
music psychologist as for the music analyst. It is inevitable that it will be compared -
in terms of its analytical efficacy, and its promise as a source for experimental
hypotheses - with the only other theory of music cognition that is as
comprehensively developed and expressed, that of Lerdahl and Jackendoff (1983).
In contrast to Lerdahl and Jackendoff, whose theory sets out to account for “the
intuitions of a listener experienced in a given idiom”, Narmour sets out to model -
and to account for the nature of - musical experience itself. As he puts it (ACBMS,
p278), “The theory is intended to capture the everyday experience of competent
music listeners”. Hence, features that Lerdahl and Jackendoff represent within
separable analytical frameworks, such as relative structural stability of events
(represented within the time-span reduction) and the dynamics of tension-relaxation
schemata (modelled within prolongational reduction) are integrated and co-
articulated within Narmour’s theory to provide an exhaustive account of the
ongoing experience of music. Thus direct comparison of their theories, and of the
predictions of their theories is difficult, and can be conducted explicitly in only a few
domains, one of these being grouping; such a comparison must, however, take into
account something of this difference in ostensible aims between the theories.

In Narmour’s theory, a pitch may be shared between two structures, functioning as

the end of one and the beginning of the next simultaneously. It will thus be
transformed (translated) on to another hierarchical level, functioning at a higher
level than that of the musical surface. This is quite different from Lerdahl and
Jackendoff’s conception that a pitch will function either in one group or another but
not in both unless the second group is contained in the first group (Grouping Well-
Formedness Rule 4, “If a group G1 contains part of a group G2, it must contain all of
G2.”), although they introduce a rule-modification to allow for overlaps or elisions
(Grouping Overlap principle, see Lerdahl and Jackendoff, p60) which limits the
interaction between successive groups to the sharing of one note that has no

16
implications for, e.g., the relationships between the notes on either side of the
shared note, such notes being capable of playing roles in Narmour’s theory (in
respect of his “tangled hierarchies”). This betokens a quite different conception of
note-to-note relations, though such a “note of elision” (in Lerdahl and Jackendoff’s
theory) would fulfil a similar function to a shared note between successive style
shapes in Narmour’s theory in functioning at a hierarchical level higher than that of
the musical surface. In a sense, Narmour’s treatment of separation is in direct
contrast to Lerdahl and Jackendoff’s treatment of overlapping. They treat
overlapping within grouping as something that is uncommon and exceptional, the
rule being separation, whereas Narmour treats overlap as the norm, the exception
being separation. In this respect, Narmour’s theory encapsulates ongoingness at
multiple levels to a much more fundamental degree than does that of Lerdahl and
Jackendoff. Despite this difference, the higher-level structures that result from the
application of Narmour’s theory often look quite similar to those that might result
from the application of Lerdahl and Jackendoff’s theory.

There are, however, many differences in the conceptual function of basic musical
entities and phenomena between the theories. For instance, for Lerdahl and
Jackendoff a large melodic leap is likely to function disjunctively, articulating
separation of groups. For Narmour, such a leap is likely to give rise to a “reversal
implication”, essentially fulfilling a function in the “ongoingness” of the music rather
than interrupting its flow. In fact, the position that the intervallic leap holds in
Narmour’s theory is a little difficult to square with its role as empirically
determinable within the Gestalt approach (see, for example, Bregman, 1990), which
Narmour claims lies behind the operation of his theory. In terms of auditory stream
segregation or auditory scene analysis, a large pitch interval occurring in the context
of small pitch steps is more likely to function disjunctively - as indicating a
separation of streams - than conjunctively. Perhaps Narmour’s idea of a large
interval as implying “change of registral direction and a sequence of intervallic
differentiation” is better conceived of as being derived from the examination of
Western Classical musical style structures rather than from any specific and innate
properties of our cognitive systems. However, it seems to me that Narmour’s
position could be squared with the empirical evidence by postulating that a large
interval should be generally construed as implying a change of tessitura and hence a
disjunction but, if unexpected (i.e. not part of an ongoing process) the original
implication of the interrupted process could be held to persist in the domain of
tessitura so as to imply return in the direction of the same pitch range (a reversal), at
least over the short term. If such a return does not come about, then the large
interval could be construed as a retrospective disjunction. This redefinition would
not involve any great change in the theory, given that all its elements - post-denial
implicative persistence etc - are already incorporated in the model.

One other point of necessary comparison between the theories is the degree to
which they can account for moment-to-moment musical experience. In fact, Lerdahl
and Jackendoff’s original (1983) theory does not attempt this, but Jackendoff’s recent
(1991) proposals for real-time parsing present a direct competitor to Narmour’s
theory. Again, however, comparison between Jackendoff's “parallel multiple-
analysis model” and Narmour's “implication-realisation” model is difficult to make,
but this time for the reason that Jackendoff is not nearly so explicit as Narmour in
outlining the fine detail of his theory, which seems to me to require integration with
the recent theories of Lerdahl (particularly with that of Lerdahl, 1988) to achieve
sufficient explicitness. In fact, the theory that Jackendoff sketches out is quite close,
in many of its elements - e.g., in allowing both prospective and retrospective

17
interpretations - to that of Narmour. Moreover, Narmour has certainly dealt with
Jackendoff’s objection to the “difficulty with basing a theory of musical affect on the
standard intuitive notion of expectation”, by producing an extremely complex
theory that is very far away from the “standard intuitive notion of expectation” that
Jackendoff decries, and by allowing sufficient richness in outlining his structural
principles to ensure that re-hearing is likely to maintain level of arousal, as well as
making explicit provision for this maintenance of arousal by the “reflexive” and thus
involuntary operation of his fundamental principles. However, Jackendoff's theory,
with its complex dialectic between the abstract structures embodied in our cognitive
systems and the consciousness to which they are affectively - indirectly - linked
might ultimately present a more appropriate dichotomous framework within which
to articulate the dynamics of ongoing musical perception than does Narmour's
perhaps over-rigid opposition between top-down and bottom-up processing.

Despite these caveats, Narmour has produced yet another vastly impressive work.
His theory seems to be in a continual state of development (as witness not only the
more advanced subject-matter of this volume but also the differences in emphasis
with its predecessor), and further volumes, dealing with aspects of potential conflict
between low and high-level patterning, the nature of harmonic prolongation, and
the effect of counterpoint on melodic structure (among other topics), are promised.
His productivity and attention to detail inspire awe, and if this review has seemed
over-critical it is only because his theory’s scrupulous comprehensiveness affords
the reader the luxury of indulging in informed engagement with its ramifications.
Even if his theory is eventually shown to have no relation to musical perception, he
has still produced a method of melodic taxonomy that is Linnaean in its scope. In
the current state of knowledge, his theory provides what may be the best - and is
certainly the most complete - explanation of the concatenation and accumulation of
the tiny fluxes of what might be termed “melodic energy” that make melodies
worth listening to.

18
 References

Blacking, J. (1981). The problem of ‘ethnic’ perceptions in the semiotics of music. In

W. Steiner (Ed.), The Sign in Music and Literature. University of Texas Press:
Austin.
Bregman, A. S. (1990). Auditory Scene Analysis. M.I.T. Press: Cambridge, MA.
Cross, I. and Stobart, H. (in preparation). Rhythmic structure in non-Western
culture: perception and performance.
Cross, I., Howell, P. and West, R. (1983). Preferences for scale structure in melodic
sequences. Journal of Experimental Psychology: Human Perception and
Performance, 9 (3), 444-460.
Jackendoff, R. (1991). Languages of the Mind. M.I.T. Press: Cambridge, MA.
Jones, M. R. and Yee, W. (1993). Attending to auditory events: the role of temporal
organisation. In S. McAdams and E. Bigand (Eds.) Thinking in Sound. Oxford
University Press: Oxford.
Krumhansl, C. L. (1990). Cognitive Foundations of Musical Pitch. Oxford
University Press: Oxford.
Lamont, A. and Cross, I. (1994). Children’s cognitive representations of musical
pitch. Music Perception, 12 (1), 27-55.
Lerdahl, F. (1988). Tonal pitch space. Music Perception, 5, 315-350.
Lerdahl, F. and Jackendoff, R. (1983). A Generative Theory of Tonal Music. M.I.T.
Press: Cambridge, MA.
Narmour, E. (1977). Beyond Schenkerism. University of Chicago Press: Chicago.
Narmour, E. (1989). The Analysis and Cognition of Basic Melodic Structures.
University of Chicago Press: Chicago.
Parncutt, R. (1989). Harmony: a Psychoacoustical Approach. Springer-Verlag:
Berlin.
Rose, S. (1993). The Making of Memory. Bantam Press: London.
Salzer, F. (1962) Structural Hearing. Dover: London.
Sloboda, J. A. (1985). The Musical Mind. Oxford University Press: Oxford.
Sloboda, J. A. and Parker, D. H. (1985) Immediate recall of melodies. In Howell, P.,
Cross, I. and West, R. (Eds.) Musical Structure and Cognition. Academic Press:
London.
Stobart, H. and Cross, I. (1994). Aspects of rhythmic structure and perception in the
music of Northern Potosí, Bolivia. Paper presented at the 3rd International
Conference on Music Perception and Cognition, Liège, July 1994.
Trehub , S. E and Trainor, L. J. (1993) Listening strategies in infancy: the roots of
music and language development. In S. McAdams and E. Bigand (Eds.) Thinking
in Sound. Oxford University Press: Oxford.

Ian Cross
Cambridge
June 1994

19