In music patterns are everywhere. We can find figural pitch patterns such as the Alberti bass, harmonic patterns such as the authentic cadence and along with these so well known examples to deserve a name on their own, there are plenty of repeated schemes and recognizable elements filling the score, the performance and the listening activity on different levels of the skilled audience. Yet if we consider the use of the word “pattern” in music we may find it minimally or at least differently articulated than what we might expect. For example, DeLone’s book “Music-Patterns and Style”, one of the few music textbooks that mention the word “pattern” in its title, doesn’t seem to use it extensively within the text, or even to refer to it whilst developing the subject. There isn’t even a single entry for “pattern” in the gigantic “Grove Dictionary of Music and Musicians”. And this is quite normal, as the word “pattern” is a word of common and wide use in English. Reason why it’s easy to find within texts that take its meaning for granted and avoid it’s definition or that use it as a simple member of a set of synonyms.
Nevertheless we can find an explicit use of “pattern” as a technical word in a variety of disciplines ranging from science or art to craftsmanship, such as tailoring, software design, architecture, finance, trading, knitting, interface design, decoration, data mining, machine learning, medical diagnosis and many more.
There is one field of research whose name does include the word “pattern”, “pattern recognition”, whose topics may range from formal languages to statistical decision theory, fuzzy logic, digital hardware, pure math, computer programming and so on” (see Bezdek 1999, 1). Yet, it is very difficult to find an explicit definition of “pattern” in the texts dealing with its recognition so we’ll need to turn to the definition of “pattern recognition” as a general area of study.
According to various standard definitions, “pattern recognition [is] the act of taking in raw data and taking an action based on the ‘category’ of the pattern” (Duda 2001, 4) or more precisely “the automatic discovery of regularities in data through the use of computer algorithms and with the use of these regularities to take actions such as classifying the data into different categories” (Bishop 2006, 1).
An uncommon definition of “pattern” in a text dedicated to “pattern recognition” refers to its etymological meaning, which is important to report here: “the word pattern is derived from the same root as the word patron” (a Middle English word taken from the Old French) “and, in its original use, means something which is set up as a perfect example to be imitated. Thus pattern recognition means the identification of the ideal which a given object was made after” (Pavlidis 1980 1).
A “pattern” is the result of finding some kind of regularities, that can be drawn from a set of models or “ideals”, within some sort of ready-made data. This is a phenomenon that has to already be there in the first place, in the reduced form of raw data, ready to be picked up by the recognition process. Sometimes (see Mumford 2010,1) the input data is characterized as “signal”, so that the whole process might be seen within the perspective of a “theory of signals” and the presence of some sort of “sender” is put into play, even if it’s role in driving the process is normally not considered.
Recognizing implies the action of linking a reference model to a part of the raw data, that in this way is distinguished from the “background noise” as the instantiation of the appropriate model. This needs to be already present in some way as a sort of “previous knowledge” that is already implemented in the actor either as a set of primitives, such as the so-called “geons” (Biederman 1987), or as a set of stochastic or fuzzy techniques or eventually as variables that have been set during a training stage driven by humans (as in the case of neural nets).
“Pattern recognition” is an active and complex process that can include operations such as segmentation, feature extraction and clustering, but, as Dennett (Dennett 1991) has shown, it can ultimately be described as a compression of the raw data’s description into a new and more efficient one, to become at least shorter than the simple listing of the original raw data. In this respect “pattern recognition” can be seen as a “search for structure in data” (Bezdek 1981, 1).
A variation in the use of the word “pattern” can be found within the context of Design. While “Pattern Recognition” together with its side discipline “Pattern Theory” may pursue as a hidden agenda the objective to “gain deeper understanding and appreciation for pattern recognition systems in the natural world – most particularly in humans” (Duda 2001, 4), within the field of Design there are much more specific needs to be accounted for, which are related to function and production (Munari 1968, 358) and lead to statements of purpose within the definition itself.
A “design pattern “describes a particular recurring design problem that arises in specific design contexts, and presents a well-proven generic scheme for its solution” (Bushman 1996, 8). Moreover it’s intended to “make object-oriented designs more flexible, elegant, and ultimately reusable (Gamma 1998, introduction 1). In sum, “each pattern describes a problem which occurs over and over again in our environment, and then describes the core of the solution to that problem, in such a way that you can use this solution a million times over, without ever doing it the same way twice” (Alexander 1977). And this is true not only for ”computer application design”, but for all the variety of design directions, from architecture to landscape design, to social organization up to the abstract layers of knowledge patterns and ontology patterns. A final more concise and yet illuminating definition describes “design patterns” as “structured, packaged problem solutions in literary form” (Evitts 2000 8).
What in “pattern recognition” was considered as simple raw data is here characterized as a “design problem”. This is an organized set, or better, a “domain” of elements which need to be properly arranged in order to reach an acceptable configuration. A “design pattern” process starts when a contingent problem, which is a configuration to be reached in a certain domain (for example, the behaviour of an application), is recognized as an instance of a more general representation of the problem (the pattern presented in a descriptive form related to another domain), which needs to be instantiated in the domain of the solution (for example the object oriented programming code), given that a causal connection can be made between the configuration in the domain of solutions and the configuration in the domain of the problem.
While “pattern recognition” is about linking a raw data phenomenon to a model, “design patterns” are intentional patterns linking different domains included in the meta-domain of a problem-solution structure. They are “techniques”, sets of instructions to follow in order to reach a goal in a given domain (an application, a new building, a new bank management). They are schemes, as fraud-schemes (such as the Ponzi-scheme) or as military strategies (such as in the Chinese book of “36 Stratagems”) or as any kind of repertoire of instructions that serve a purpose either in a single domain (the Ponzi- scheme is meant to accumulate money and can hardly be used in different contexts) or in a variety of very different domains (the 36 stratagems can be used in a vast range of human interactions).
The word “pattern” is also widely used within the disciplines of art, craftsmanship and music. According to the Oxford English Dictionary a pattern is (1) “a repeated decorative design”, (2) “a regular and intelligible form or sequence discernible in the way in which something happens or is done”, (2) “a model or design used as a guide in needlework and other crafts” or (4) “an excellent example for others to follow”. The first definition already links the word “pattern” to decoration and indeed “pattern” plays a technical role in visual arts. Nevertheless in music, “pattern” doesn’t seem to be used as a technical word at least for what concerns most pedagogic and musicological texts. In spite of the huge relevance of the role of repetition in music, as Ockelford showed with numerous quotes in the introduction of his text about “Repetition in Music” (Ockelford 2005), the word “pattern” is utterly unused when talking about music outside its general meaning within the ordinary language. There is a curious ambiguous sense already in the Oxford Dictionary’s definition, which on the one hand points to an action in the past (the design was repeated by somebody), whilst on the other hand it points to the use of the past participle as an adjective (the property of the design is being “repeated”). In the duplicity of this kind of “pattern”, that we could call an “expressive pattern”, as something which is intentionally made to be perceived as such, lies one of the most important features of this use of the word “pattern”.
This duplicity leads to the distinction of two domains which we’ll name after the semiotics of Jean Molino (Nattiez 1987): the “poietic domain” and the “aesthesic domain”. The first one is related to the practices of an “actor” whom creates the pattern on a particular medium, the score in notated music. The music performance of that score will be sensorially received by another actor whom will possibly recognize the poietic pattern in the “aesthesic domain”, namely the domain in which listening distinctions take place. The two domains are definitely separated from each other, as poietic patterns – which in music are related to gestures, notation and composition techniques in general – might be very different, also in time organization, from aesthesic patterns recognized when listening: sometimes a poietic pattern might not even trigger a recognition in the aesthesic domain, like it happens frequently with dodecaphonic series repetitions. In the middle there is an object in its proper domain, the so-called “neutral level”, such as a score, which is subject to possible transformations by the intermediacy of human or automated actors (such as musicians, audio compression, sound distribution and so on).
There is here a strong constraint implied by the idea of repetition: a poietic/aesthesic configuration must be repeated sequentially. Or better: a pattern is something that is repeated sequentially. For this reason and unlike the previously seen characterizations of “pattern”, “expressive patterns” are not abstract models existing in domains different from their instance ones. On the opposite: they have the same, so to say, ontological status as their copies: a pattern is recognized only if its enunciation is immediately followed by its iterations and the first enunciation may achieve the status of pattern only because it is the first one to be enunciated. This is true because the following copies might not be perfect reproductions of what we could call the original “pattern” (for instance, the “Alberti bass” pattern needs to be recognizable even when harmony changes), so there is no other reason to choose the first enunciation as a kind of an ideal model rather than its precedence in time ordering.
Another important issue is the emphasis put on a hierarchical structure: we recognize a pattern only when we face multiple instantiations of that pattern at the same time, so that a pattern is immediately related to a larger element, that exhibits a perceivable unity. We can finally come to what I think is the essence of the use of the word “pattern” in music or in general of what we call an “expressive pattern”: a pattern is a configuration in a domain linking to configurations that follow it and are recognizably similar to it, in a contiguous time.
If repetitions are so important in music, it’s quite consequent to wonder why the word “pattern” is so underrated as a technical word in music. The answer is pretty simple: it’s because repetitions of different elements and with different features have been called with various names along the history of western music, following the evolution of styles, genres, composition practices and so forth. We have refrains, ostinatos, subjects, passacaglias, forms, themes, chord progressions and many many more items. And the more names there, the more difficult it is to be aware of the common underlying principle of repetition, which seems to be hidden in plain sight.
No wonder then if Ockelford, while realizing that repetition is a relation among music fragments, nevertheless feels the need to invent a new complicate vocabulary, without even mentioning the word “pattern”. So when he comes to define the relationship between two similar music fragments as “the interperspective relationship through which imitative order is perceived” he ends up naming it a “zygonic” one (Ockelford 2005, 21).
We can find at least a fourth use of the word “pattern”, which is still related to music, but mostly lies outside the academic studies of western, so-called, classical tradition. In an exemplar definition taken from a handbook of Jazz exercises, patterns are “finger habits. (…) Something that is not necessarily pre-heard, but (…) is understood to work (by cognisanze of the theoretical reasons and/or by previous experience), or (…) that feel comfortable to the fingers and hands. [Something] (…) to be practiced diligently by serious students” (Coker 1970, 1).
In a wide range of handbooks dedicated to the pedagogy of popular music improvisation, “patterns” are music fragments that must be repeated and transformed (mostly by pitch transposition) on and on so that it can become instinctive to perform them in a variety of occasions.
A “training pattern” is a configuration in a given domain, for example a fragment of music notation, that is instantiated in the performative domain within the context of training. What is important here seems to be the improvement, related to the process of instantiation, in managing what we have called the “previous knowledge”.
In machine “pattern recognition” this knowledge is provided by humans in form of machine implementation. In “design patterns” that knowledge is presented as descriptions (“literary” ones as Evitts says). In “expressive patterns” it is present as well-learned repetitive practices both in the poietic and aesthesic domains (once again with the help of some “literary” instruction).
Here, the “previous knowledge” is progressively hard-wired as a behaviour, becoming a “habit”, so that in the end when the improviser is instantiating a pattern, it somehow disappears as an instance among instances. Only the memory of its first appearance remains, as a label in a different domain, in a transfiguration in which we might see the emergence of meaning.
We have seen four paradigmatic uses of the word “pattern” by focusing on elements that, while being present in some way in all the uses of “pattern”, are emphasized in different ways within each definition. These elements are: recognition, similarity, repetition, domain, hierarchy, previous knowledge, type of relation/structure and habit.
The presence of a structure in the relationship between domains may suggest that in fact we can see the four uses of the word “pattern” as patterns themselves.
Let’s go back to the beginning, with a last definition:
“Pattern Recognition is the scientific discipline dealing with methods for object description and classification” (De Sà 2001, 2).
This reminds us that whilst talking about processes and relations, we still largely rely in dealing with a world of objects, which might be emerging in our language due to repetitive practices such as looking for resemblances, not too differently from the way we listen to music.
Alexander C, Ishikawa S, Silverstein M (1977), A Pattern Language: Towns, Buildings, Construction, Oxford University Press.
Bezdek J (1981), Pattern Recognition with Fuzzy Objective Function Algorithms, Plenum,
Bezdek J, Keller J, Krishnapuram R, Pal N. (1999), Fuzzy Models and Algorithms for Pattern Recognition and Image Processing, Kluwer Academic Publishers Norwell, MA.
Biederman I (1987), Recognition-by-components: a theory of human image understanding, in Psychol Rev. 94(2) pp. 115-47.
Bishop C (2006), Pattern Recognition and Machine Learning (Information Science and Statistics), Springer-Verlag, NY.
Bushman F & al, (1996), Pattern-Oriented Software Architecture. Volume I. A System of Patterns, John Wiley and Sons, NY.
Coker J, Casale J, Campbell G, Greene J (1970), Patterns for Jazz, Studio P/R.
DeLone RP (1971), Music: Patterns and Style, Addison-Wesley, Reading, MA.
Evitts P (2000), A UML Pattern Language, Macmillan Technical Publishing, Indianapolis Gamma
E. Helm R. Johnson R. Vlissides J (1998), Design Patterns, Elements of Reusable Object Oriented Software, John Wiley and Sons, NY.
Dennett D C (1991), Real Patterns, in The Journal of Philosophy, 88 (1) pp. 27-51.
De Sà M (2001), Pattern Recognition – Concepts, Methods And Applications, Springer-Verlag, NY. Duda R, Hart P, Stork D (2001), Pattern Classification, John Wiley and Sons, NY.
Munari B (1968), Design e comunicazione visiva, Laterza, Bari.
Mumford D & Desolneux A (2010), Pattern Theory: The Stochastic Analysis of Real-World Signals. A K Peters, Natick, Mass.
Nattiez J (1987), Musicologie générale et sémiologie. [it. transl. (1989) Musicologia generale e semiologia. EDT, Torino.]
Ockelford A (2005), Repetition in music: theoretical and metatheoretical perspectives, Ashgate, London.
Pavlidis T (1980), Structural Pattern Recognition, Springer-Verlag, NY.