TYPE Review PUBLISHED 18 November 2022 DOI 10.3389/fpsyg.2022.1033561 Structural analyses in the study OPEN ACCESS of behavior: From rodents to non-human primates EDITED BY Jorge L. Puga, University of Granada, Spain REVIEWED BY Maurizio Casarrubea 1*, Jean-Baptiste Leca 2,3, Noëlle Gunst 2, Miguel Pic, South Ural State University, Russia Gudberg K. Jonsson 4, Mariona Portell 5, Giuseppe Di Michael Brill, Giovanni 6, Stefania Aiello 1 and Giuseppe Crescimanno 1 Julius Maximilian University of Würzburg, Germany 1 Laboratory of Behavioral Physiology, Department of Biomedicine, Neuroscience and Advanced Diagnostics (Bi.N.D.), Human Physiology Section “Giuseppe Pagano”, University of Palermo, *CORRESPONDENCE Palermo, Italy, 2 Department of Psychology, University of Lethbridge, Lethbridge, AB, Canada, Maurizio Casarrubea 3 School of Natural and Engineering Sciences, National Institute of Advanced Studies, Bangalore, maurizio.casarrubea@unipa.it India, 4 Human Behavior Laboratory, School of Health Sciences, University of Iceland, Reykjavík, SPECIALTY SECTION Iceland, 5 Department of Psychobiology and Methodology of Health Sciences, Universitat This article was submitted to Autònoma de Barcelona, Cerdanyola del Vallès, Spain, 6 Laboratory of Neurophysiology, Quantitative Psychology and Measurement, Department of Physiology and Biochemistry, Faculty of Medicine and Surgery, University of Malta, a section of the journal Msida, Malta Frontiers in Psychology RECEIVED 31August 2022 ACCEPTED 25 October 2022 The term “structure” indicates a set of components that, in relation to each other, PUBLISHED 18 November 2022 shape an organic complex. Such a complex takes on essential connotations CITATION of functionally unitary entity resulting from the mutual relationships of its Casarrubea M, Leca J-B, Gunst N, constituent elements. In a broader sense, we can use the word “structure” to Jonsson GK, Portell M, Di Giovanni G, Aiello S and Crescimanno G (2022) define the set of relationships among the elements of an emergent system Structural analyses in the study of behavior: that is not determined by the mere algebraic sum of these elements, but by the From rodents to non-human primates. interdependence relationships of these components from which the function Front. Psychol. 13:1033561. doi: 10.3389/fpsyg.2022.1033561 of the entire structure itself derives. The behavior of an integrated living being COPYRIGHT can be described in structural terms via an ethogram, defined as an itemized © 2022 Casarrubea, Leca, Gunst, Jonsson, list of behavioral units. Akin to an architectural structure, a behavioral structure Portell, Di Giovanni, Aiello and arises from the reciprocal relationships that the individual units of behavior Crescimanno. This is an open-access article distributed under the terms of the establish. Like an architectural structure, the function of the resulting behaving Creative Commons Attribution License (CC complex emerges from the relationships of the parts. Hence, studying behavior BY). The use, distribution or reproduction in in its wholeness necessitates not only the identification of its constitutive units other forums is permitted, provided the original author(s) and the copyright in their autarchic individuality, but also, and importantly, some understanding owner(s) are credited and that the original of their relationships. This paper aimed to critically review different methods publication in this journal is cited, in accordance with accepted academic to study behavior in structural terms. First, we emphasized the utilization of practice. No use, distribution or T-pattern analysis, i.e., one of the most effective and reliable tools to provide reproduction is permitted which does not structural information on behavior. Second, we discussed the application of comply with these terms. other methodological approaches that are based on the analysis of transition matrices, such as hierarchical clustering, stochastic analyses, and adjusted residuals. Unlike T-pattern analysis, these methods allow researchers to explore behavioral structure beyond its temporal characteristics and through other relational constraints. After an overview of how these methods are used in the study of animal behavior, from rodents to non-human primates, we discussed the specificities, advantages and challenges of each approach. This paper could represent a useful background for all scientists who intend to study behavior both quantitatively and structurally, that is in terms of the reciprocal relationships that the various units of a given behavioral repertoire normally weave together. Frontiers in Psychology 01 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 KEYWORDS hierarchical clustering, transition probabilities, adjusted residuals, T-pattern analysis, behavioral structure–function interface Introduction rooms, or an interior space from the exterior of the house. The wall, then, does create a division that makes possible a better use of two rooms that would otherwise be a single, much less usable “The ability to reduce everything to simple fundamental laws space. However, this explanation says nothing about how the wall does not imply the ability to start from those laws and came to be. As much as the answer just given regarding the reconstruct the universe.” finalistic aspects is more or less articulated, in fact, we have left out Anderson (1972) the second and essential part of the above binomial: the processual component. In contrast to function, the concept of “process” is This paper deals with behavior and with a number of imbued with purely mechanistic connotations. The term “process” approaches useful to study it. We asked: “what is the behavior of a therefore indicates how something happens. The fundamental living being?” Such a question, simple and straightforward, cannot question related to it is no longer “what does it do?” but “how does be propaedeutic to an answer characterized by the same features. it do it?.” Returning to our example, then, we may say that a wall, Using a common scientific database and searching the term capable of performing the above functions, consists of the “behavior” produce thousands of definitions, explanations, points juxtaposition of bricks, with a specific orientation within the of view, theories etc. Some definitions are very cryptic, others apartment, properly joined together thanks to a specific amalgam fanciful, others difficult to understand and others more or less that prevents displacements between neighboring bricks and gives agreeable. Not surprisingly, some of the strongest definitions of greater stability to the structure. As will be noted, these purely behavior were proposed by the founding Figures of Ethology such mechanistic aspects complement and explicate the question about as Konrad Lorenz, Nikolaas Tinbergen or Irenäus Eibl-Eibesfeldt: the above purpose. behavior is the reaction of a living being to external (Tinbergen, Even the unexperienced reader will not miss the enormous 1951a) or internal (Tinbergen, 1951b) causal factors and it is difference between the concepts of “function” and “process” and organized on the basis of a sequence of events in time (Eibl- their very close connection with the structure generating them. Eibesfeldt, 1970). These words echo a central dogma in ethology All of this, while of wide application in numerous fields (such as and in most disciplines belonging to behavioral sciences. Indeed, the example of wall and bricks), finds extraordinary congruity, in they relate to the causality and temporal organization of behavior. the biological realm, with one discipline in particular. The One thing is self-evident: like its definition, the methods to study concepts of “function” and “process” permeate Physiology to the behavior are not simple and straightforward. extent that they constitute the deeper and essential “raison d’être” of this entire discipline. In Physiology, these two concepts are the two sides of the same coin. The examples are potentially endless Structures, functions, and processes and all, conceptually, similar to the example of the bricks and the wall. For instance, when studying the heart, physiologists typically In his well-known writing titled “Principles of the self- describe the various properties of this extraordinary organ, organizing system,” the brilliant English psychiatrist William Ross pointing out how its prodigious morphological characteristics Ashby stated that if the relationship between two entities, e.g., “A” (i.e., structure) enable it to constantly and tirelessly feed blood into and “B,” becomes conditional on “C,” then a necessary component the aorta and pulmonary vessels (i.e., functions) through specific of organization is present and, as a result, a whole composed of mechanisms involving both electrical and purely mechanical interacting parts (i.e., a structure) emerges (Ashby, 1962). Closely phenomena (i.e., processes). Everything in the domain of related to the concept of “structure” is the dual and dynamic Physiology orbits around functions and processes, and arguably, relationship between function and process. The concept of there is no system, organ, tissue and/or even single cell, able to “function” is permeated with purely finalistic, i.e., teleological, escape such a binomial. Studying Physiology means, therefore, connotations. Thus, the word “function” denotes the purpose of being able to describe with absolute precision each of the two sides the object “sensu lato,” whether it is a very simple or extremely of this coin: the functions and processes, the why and the how. complex item. Two common questions, that is “what is it for?” and These concepts apply to the study of behavior for the reasons “what is its purpose?,” perfectly express the teleological nature of discussed below. any object. What is the purpose of one of the many walls we can Detailed knowledge about behavioral structure contributes to easily observe in a house? While trivial, this example gives a good testing hypotheses about behavioral function. The “design-feature understanding of the concept of teleology: it serves to separate two argument” holds that thorough structural analysis of a given Frontiers in Psychology 02 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 behavior pattern provides valid information about its hypothesized sub-units, use simple and unambiguous names to label them, and function (Martin and Caro, 1985; Moran, 1985). The heuristic generate operational definitions to describe them, thereby power of this behavioral structure–function interface is reflected constructing an “ethogram.” For example, the behavior of a rodent in the following statement by Pellis and Pellis (1998): “Therefore, may be referred to as “Walking” when the animal moves relatively behavioral description informs functional inference, which in turn, slowly and quadrupedally from one quadrant of the arena to influences further description” (p. 115). Thus, broad similarities another, “Climbing” when its upper body leans against the and subtle differences in the structural organization of perimeter walls, “Rearing” when it lifts up on its hind legs without evolutionarily related behaviors are indicative of their respective leaning against any of the nearby walls, “Face Grooming” or “Body motivational underpinnings and functional features. This Grooming” when it picks at or scratches its head or other body approach has proved particularly useful to compare pairs of parts, respectively, with its paws or mouth, etc. Identifying each of behavioral traits that are developmentally and evolutionary linked, these behavioral units is a relatively simple task that only requires but vary in their functional constraints. For example, among a fairly short training period. Practically, the observer watches the different types of object manipulation in non-human primates and video of the animal’s behavior and, through a specific computer pre-school children, researchers used structural variables, either program, e.g., The Observer software coder (Noldus Information based on kinematic or temporal components, to infer underlying Technology, Netherlands), scores the occurrences of each psychological mechanisms and explain the actions being behavioral unit. The product of this simple, but terribly time- performed in terms of relative purpose and utility (e.g., object play consuming, process is called event log-file and consists of a and object exploration: Hughes, 1978, 1979; percussive object play generally long list of the behavioral events performed by the and extractive foraging: Pelletier et al., 2017; Pellis et al., 2019). animal with their respective onset times. An event log-file can These studies indicate that utilitarian motivational processes and be conveniently illustrated by means of an event plot. Figure 1 is functional constraints in object manipulation covary with an example of an event plot obtained from one, specific pathogen- behavioral structure. When comparing two types of manipulative free, 2-month-old male C57/BL wild-type mouse (Jackson activities, the more product-oriented one (e.g., object exploration, Laboratories, United States), belonging to a group of subjects, extractive foraging) show higher levels of kinematic and/or observed in the open-field for 10 min (Casarrubea et al., 2020b). temporal structure than the more process-oriented one (e.g., A simple question arises: after watching 10 min of this video, after object play; see also Rasa, 1984). carefully observing the rodent’s exploratory behavior through the In line with one of the basic tenets of Darwinian evolution, open field, after recognizing hundreds of Walking, Climbing, holding that selection strength and phenotypic variability are Rearing, Face Grooming etc., what to do with these data? Figure 1 negatively correlated, more functionally constrained behaviors are shows more than 300 behavioral units. In abscissa we report the subjected to higher selection pressures, which in turn, lead to less time, in ordinate the behavioral unit performed. The plot structurally variable behaviors. Thus, the behavioral structure of highlights where each behavioral unit falls in time. Walking object-directed activities may be used as a proxy to assess their occurred 114 times, Rearing 9 times, Immobile-Sniffing 136 times relative functionality. Theoretically, this powerful structure– etc. Table 1 shows the name, frequency, percentage and overall function interface could be applied to other pairs of behavioral duration of the behavioral units plotted in Figure 1. This table can traits that are linked at the proximate levels (i.e., in their be enriched with numerous additional quantitative evaluations developmental trajectories and underlying sensori-motor and such as, for example, the average duration of each behavior, the cognitive mechanisms) and at the ultimate levels (i.e., in their latency of the first appearance of each behavior etc.; one may also functional consequences and phylogenetic pathways; Tinbergen, do somewhat more complex analysis such as, for example, 1963). Among other pairs of behavioral traits that were subjected evaluating the average duration of each item in relation to the time to differential functional constraints and whose putative of observation, so as to provide a time course for each minute of mechanistic and evolutionary connections are unravelled by observation. The first evidence, even with a very first glance of in-depth structural analysis, let us mention object play and tool Table 1, is the remarkable feeling of completeness and use (Cenni et al., 2020, 2022) as well as female-to-female mounting comprehensiveness. The catch is, however, just around the corner. and female-to-male mounting (Gunst et al. 2020, 2022). All these numbers per se tell us little about the complexity of an individual’s behavior. Taking into consideration our trivial example with the wall and the bricks, we argue that describing Physiology of behavior behavior in terms of individual units is like considering individual bricks and their many characteristics (i.e., weight, height, length, Behavior, whether of an insect or a primate, can be broken width etc.); however, it ignores the most important aspect, namely, down into simpler units, each indicative of a specific and what flows from their relationships both in terms of structure and characteristic part of the subject’s behavioral repertoire. Imagine the related functional and processual elements. As we previously a fruit fly flying in a room, a rat moving through an Open Field, a noted, using a different and perhaps more effective metaphorical non-human primate exploring a new part of a forest. Ethologists image, evaluating a subject’s behavior by considering only the typically divide each of these activities into several behavioral individual units and not their relationships “[…] is not different Frontiers in Psychology 03 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 FIGURE 1 Event plot of 8 behavioral units (Y-axis) occurring during a 10-min (X-axis) observation window. For abbreviations of behavioral units on Y-axis see Table 1. Data from one subject taken from Casarrubea et al. (2020b). See text for details. TABLE 1 Synoptic table summarizing occurrences, percent purpose. The question, then, is about where the meaning of distribution and duration of behavioral units listed in the first column behavior lies. Is it in the study of isolated behavioral units, each (abbreviations in brackets) and illustrated by means of event plot in Figure 1. characterized by dozens of numbers, but still fragments of what is the animal’s true behavior? Or, conversely, is it in trying to relate Component Occurrences Percent (%) Duration (sec) all the behavioral units described in our observation in an attempt Walking (Wa) 114 37.50 181.57 to reconstruct the true behavior? The answer is obvious. Studying Rearing (Re) 9 2.96 7.63 behavior in its wholeness therefore means relating the individual Immobile-sniffing (IS) 136 44.74 329.25 units because only from these relationships arises the structure Immobility (Im) 2 0.65 2.75 that is, in turn, preparatory to the functions and processes it serves Paw licking (PL) 4 1.32 8.67 and performs. In this lies the Physiology of Behavior. However, Face grooming (FG) 4 1.32 11.7 before delving into the different methods to study the relationships Climbing (Cl) 33 10.86 50.93 existing between the different units of the behavioral repertoire, it Body grooming (BG) 2 0.65 7.5 is important to clear the field from possible misunderstandings. Tot 304 100 600 Data from 1 subject taken from Casarrubea et al. (2020b). See text for details. Simple quantities and their usefulness from classifying all the single pieces of a puzzle missing the What we discussed in section “physiology of behavior” might comprehensive picture. The functional meaning of a behavior, i.e., lead to a seemingly logical but deeply flawed juxtaposition: one the study of the existing interplay between an animal and the might believe that a purely quantitative approach is somewhat context, is a picture lying in its intrinsic structural features” useless in the context of a behavioral study. We clearly do not (Casarrubea et al., 2019a). Actually, Figure 1 contains the bricks support such a biased view. The meticulous description of each of our wall. Each dot in the plot highlights the specific “brick” unit of behavior, albeit unconnected with the relational dynamics shown on the Y axis and the time in which it occurred, shown on of the different units, can find numerous applications. If, for the X axis. Not all bricks are the same, just as the different types of example, we look again at Table 1, it is certainly useful to know the adhesives we can use to build the given structure are not the same more frequent and the less frequent behavior, or those that last at all. However, if proper glue is not used between the individual longer and those that are shorter etc. Based on this information, it units, they will not have solid reciprocal relationships and the final becomes possible to know what the subject does and, importantly, structure will not only be more unstable and/or difficult to build, how long it takes to do it. In addition, one may examine if and how but also prone to a rapid collapse. There is more. Not only we need these items change following the administration of an independent the appropriate bricks and cement for the structure we want to variable such as the benzodiazepine Diazepam, a drug with build, the means we employ should also be adequate for the anxiolytic action, that induces a conspicuous increase in Walking Frontiers in Psychology 04 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 (Casarrubea et al., 2009b), or the administration of the beta- of computational capabilities allowed by modern personal carboline FG7142, a potent anxiety inducing molecule, causing a computers. In the following sections, we present some useful significant decrease in Walking (Casarrubea et al., 2017). The approaches to describing behavior in structural terms, that is, explanation lies in the modulation of the state of anxiety of the two through the relationships among different units of the behavioral molecules: with reduced anxiety, the animal explores the repertoire. These sections do not claim to be exhaustive regarding environment more (and therefore walks more), whereas with the totality of methods that can be employed to solve these increased anxiety the subject tends to move much more problems. Rather, we present a number of approaches that, from circumspectly (and therefore walks less). These examples highlight our 30 years of experience in the field, have proven to be more the usefulness of the descriptive/quantitative approach when than efficient and useful in this regard. These approaches can dealing with discrete variables. However, in both cases, nothing be divided into two major groups: methods based on the use of can be said about what strategies the animal deploys, how it relates transition matrices and methods based on the use of to the environment etc. In short, we still know little about exactly T-pattern analysis. what the animal is doing and precisely how it is doing it; specifically, information about the pharmacological underpinnings of the structure of its behavior in its organic entirety Transition matrices and related analyses remains scanty. The structure of a hypothetical transition matrix is illustrated in Figure 2. The utilization of transition matrices implies three Structural analyses in the study of requirements (1) the number of transitions contained in a behavior transition matrix must be at least five times the number of units of the behavioral repertoire (Spruijt and Gispen, 1984); (2), the Three prerequisites, propaedeutic to each other, are essential number of empty cells must be no more than 20% (Spruijt and to study behavior through the relationships among the different Gispen, 1984; Casarrubea et al., 2008); and (3) the number of units of the behavioral repertoire: one must (a) clarify what these subjects used should be at least three times the number of relationships are, (b) have the appropriate means to study these behavioral units (Short and Horn, 1984; Espejo and Mir, 1993). relationships, (c) use these methods to tie together the different These three requirements allow researchers to generate a number units of the behavioral repertoire before any discussion about of transitions sufficient to avoid a poorly filled matrix (Spruijt and behavioral structure can occur. Gispen, 1984). How do the different behaviors illustrated in Figure 1 and carefully described in Table 1 relate to each other? Are there Hierarchical clustering constraints of any kind applied to them? If so, what methods can Cluster analysis aims to highlight constraints among units of researchers use to describe such constraints? One way to assess the the behavioral repertoire on the basis of the overall number of relationships between a behavior and the following one is to reciprocal transitions occurring during the overall observation calculate reciprocal transition frequencies. Probabilistic period. To this purpose, a transition matrix must be transformed constraints may also be evaluated by determining the probability into a similarity matrix by using an aggregative procedure. Several of behavioral unit “A” shifting to behavioral unit “B.” One may also aggregative procedures are available. For instance, in Figure 2, the examine whether the transitions between these two units occur similarity between the behavioral unit A and B can be obtained by more or less frequently than an expected value. Furthermore, it is using the procedure described by Mos et al. (1987): essential to consider what is certainly the most important relational constraint: the temporal one. Here the question S = (α / a + α / b + β / a + β / b ) 50 becomes: Is it possible for different behavioral units to occur at fixed intervals or, at the very least, is it possible to identify specific where S is the “similarity” value between units “A” and “B,” temporal distances among different behaviors or only between Greek letters “α” and “β” respectively indicate number of some? Such temporal constraints may not be, and in fact rarely transitions from unit A to unit B and from unit B to A, lowercase are, rigidly sequential. letters “a” and “b” respectively indicate the total occurrences of A All these techniques describing, and sometimes unravelling, and B, and 50 is a normalization factor. This procedure must relationships between units of the subject’s behavioral repertoire, be repeated for each pair of cells within the matrix. The result is a belong to the realm of multivariate analyses. These methods, half matrix where each cell does not express a transition but the although not recently conceived and introduced in behavioral “vicinity” between two variables (rows and columns names). This research, have seen a growing and steady diffusion only in recent approach thus assumes that the number of transitions can decades, hand in hand with the development of, and easy access be considered a valid index of aggregation between two or more to, personal computers. Indeed, most of these analyses are very elements, a premise that should be taken into serious consideration complex and involve considerable amounts of data, the when interpreting the data. Indeed, talking about “similarity” management of which greatly benefits from the continuing growth value between two behavioral units might be misconstrued as Frontiers in Psychology 05 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 FIGURE 2 Example of transition matrix. A, B, C: names of three hypothetical FIGURE 3 behavioral units; α, number of transitions from unit A to B; γ: Hypothetical dendrogram representing similarity values (S: Y-axis) number of transitions from A to C; β: number of transitions from among behavioral units (X-axis) on the basis of a half matrix B to A; ε: number of transitions from B to C; δ: number of arranged following an aggregative clustering procedure. See text transitions from C to A; ζ: number of transitions from C to B; XA, for details. XB and XC: number of auto-transitions from A to A, B to B, and C to C, respectively; a: total occurrences of A; b: total occurrences of B; c: total occurrences of C. rodent’s behavior in the specific experimental context. Importantly, these two units being qualitatively similar. In reality, the closeness these studies also demonstrated that these three categories are expresses a high number of mutual transitions, not a extremely sensitive to the administration of independent variables, phenomenological similarity at all. For example, in rats, Edge- such as, for example, specific drugs. Sniff behavior (i.e., when the animal sniffs the edge of the hole in a test called “Hole-Board”) and Head-dip behavior (i.e., when the Probabilities of transitions animal inserts its head inside the hole), although qualitatively very A stochastic approach is useful to emphasize probabilities of different, form a stable cluster in untreated subjects (Casarrubea transitions among units of the behavioral repertoire. First, on the et al., 2009a) but highly sensitive to pharmacological manipulation basis of relative frequencies of transitions among behavioral units, of the animal’s anxiety condition (Casarrubea et al., 2009b, 2017). the transition matrix is transformed in a stochastic matrix. Therefore, the strong temporal link between these two behaviors Following this step, the final matrix must meet three criteria: (1) is highly evocative of the rodent’s anxious state. From the half- for each row, the sum of all cells should be equal to 1; (2) the value matrix obtained via this procedure, researchers should create a of each cell should be between 0 and 1; and (3) the transition dendrogram, i.e., a graphical tree representation that shows in a probability from one cell to the remaining cells should be equal to very simple and intuitive way the “closeness” between the various 1. A consistent advantage of such a matrix is that it can clusters of behaviors. A hypothetical dendrogram is presented in be graphically expressed through a pathway diagram, where Figure 3. The S values of the half-matrix are shown in the ordinate. transition probabilities among cells can be represented by means In practice, this Figure indicates that units A and B, based on the of arrows, with higher probability ranges showing as thicker number of reciprocal transitions, are linked by a high S value, arrows (Hemerik et al., 2006). Like the dendrograms briefly which is matched by the value of unit C, that occupies a more discussed in section 2.1., the use of probability diagrams is “peripheral” position. This illustrative approach has been fruitfully intuitive and emphasizes the temporal constraints between used in numerous papers (Espejo and Mir, 1993, 1994; Espejo different behavioral units, providing a unified view of the et al., 1994; Casarrubea et al., 2008, 2009a,b, 2011a, 2012, 2015c, behavioral repertoire. A hypothetical probabilistic pathway 2017). From a functional point of view, these analyses made it diagram is presented in Figure 4. In practice, it can be inferred possible to categorize the different units of behavior into clusters, from the Figure that units A and B share higher probabilities of that is, “behavioral sets” in turn indicative of a more complex mutual transitions than units B and C or units A and C. These activity contextualized to the explored environment. For example, probabilistic representations have been successfully used in in the work of Espejo and Mir (1993) as well as in that of several studies (Espejo and Mir, 1993; Espejo, 1997; Lino-de- Casarrubea et al. (2006), three different clusters of behavioral units Oliveira et al., 2005; Casarrubea et al., 2008, 2009a,b; Santangelo have been described with regard to rat behavior on the hot-plate: et al., 2018). Functionally speaking and similar to what was sniffing, primary noxious-evoked and escape responses. This mentioned in the previous section regarding dendrograms, these categorization, highlighting the animals’ response to the analyses of the relationships among behavioral units in terms of nociceptive stimulus, allowed for a better interpretation of the probabilistic constraints allowed researchers to unravel otherwise Frontiers in Psychology 06 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 FIGURE 4 Hypothetical probabilistic pathway diagram representing transition probabilities among behavioral units. Right side, the selected probability ranges used to draw probability arrows of matching thickness. See text for details. unnoticeable behavioral dynamics characterizing subjects’ activity such as lag +2 or lag-2, lag +3 or lag-3, etc. (Figure 6). Among in the specific environment. For instance, Espejo (1997), in the other examples, lag sequential analyses have been employed to Elevated Plus Maze, revealed the existence of a behavior heavily investigate the motivational underpinnings and functional polarized toward sniffing episodes and the occurrence of two components of specific behavioral elements expressed within behavioral units (Stretched Attend Posture and Protected sequences of actions performed during playful or sexual activities Dipping) sharing high reciprocal probabilities of transitions. The in free-ranging macaques (Cenni et al., 2022; Gunst et al., authors suggested that these two behavioral units may have a 2022a,b, 2022). common behavioral significance in the form of “anxiety-related behaviors, which mice usually display in succession and from Adjusted residuals relatively secure closed and central sections of the maze” (Espejo, One question that neither clustering procedures nor 1997, p. 109). probability matrices can answer is the statistical significance of the transitions between behavioral units. To this effect, Haberman Lag sequential analyses (1973) and Everitt (1977) proposed the conversion of a transition Lag sequential analyses are an example of analytical matrix into a matrix containing adjusted residuals. Adjusted techniques that rely on transition matrices (Sackett, 1979; Quera residuals are standardized residuals divided by their respective and Estany, 1984). LSA is a type of temporal analysis applied to standard deviations (Haberman, 1973; Everitt, 1977). Even though behavioral sequences that calculates the frequency of transitions a specific formula is available to calculate residuals, a dedicated between pairs of behavioral units within a certain lag. The first software allows for the automatization of this task when matrices event of the pair is called “Criterion” and the second “Target” reach hundreds cells. Various computer programs, such as (Faraone and Dorfman, 1987). Depending on which direction in Matman (Noldus Information Technology, Netherlands), make time researchers are interested in investigating (i.e., positive or the calculation of adjusted residuals significantly faster and more negative), they calculate how often the Criterion (e.g., Event A) reliable. The most important benefit of adjusted residuals is the was followed by the Target (e.g., Event B), or how often Target possibility to utilize a common Z table for their interpretation: (e.g., Event B) preceded the Criterion (e.g., Event A), respectively. values ≥ +1.96 and ≤ −1.96 indicate statistically significant There are two types of LSA, depending on the type of transitions (p < 0.05) transitions occurring more often (≥ +1.96) or less often between a Criterion and a Target. First, a “time lag” sequential (≤ −1.96) than expected. Like probability matrices, matrices of analysis requires the comparison of the same time window before adjusted residuals can be illustrated by means of pathway diagrams and after the Criterion, and considers transitions between a (Spruijt and Gispen, 1984; Spruijt, 1992; Vanderschuren et al., Criterion and a Target within these specific time windows, 1996; Van Den Berg et al., 1999), where the thickness of the arrows independent of how many other events are between them. is also indicative of the significance level of the given transition. Researchers then calculate the number of transitions from a However, pathway diagrams of adjusted residuals come with a Criterion to those Targets occurring within a specific time caveat. While representing positive residuals arrows is intuitive, window following or preceding the Criterion (Figure 5). Second, representing negative residuals is problematic as negative residuals a “state lag” sequential analysis considers the transitions between are transitions occurring significantly less often than expected, a Target that directly followed or directly preceded a Criterion making their representation as arrows counter-intuitive. One (i.e., lag +1 or lag-1, respectively). Other pairs of lag are possible, solution, in this regard, is to present the residuals by means of Frontiers in Psychology 07 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 FIGURE 5 Hypothetical example of a 30-s time lag sequential analysis along a timeline of performance of a given Criterion behavior (C) in seconds. This analysis is used to compare the frequency of a Target behavior (T) within a 30-s window before and within a 30-s window after the start of C. For instance, T can be “being approached and being looked at by close neighbors.” This example indicates that T is more frequent shortly after the performance of C than shortly before. FIGURE 6 Hypothetical example of a state lag sequential analysis used to compare the frequency of a dorso-ventral (D-V) position (i.e., Target) adopted by the female consort partner immediately before (i.e., lag-1) versus immediately after (i.e., lag +1) the performance of a female-to-male mount (i.e., Criterion: C). Other positions relative to the male consort partner during intermount intervals include dorso-dorsal (D-D) and side-by-side (S-S). Other pairs of lag are lag +2 and lag-2, lag +3 and lag-3, etc. (Gunst et al. 2022). positive and negative bars indicative of positive and negative The temporal dimension residuals, respectively. A hypothetical representation of residuals using such an approach is shown in Figure 7. Examples of such an The temporal dimension adds an additional layer of approach have been used in several studies (van Lier et al., 2003; complexity to the analysis of the possible relationships existing Casarrubea et al., 2008, 2009a,b, 2012). between different units of the behavioral repertoire. Indeed, such Adjusted residuals are extremely sensitive to changes in the relationships may not be linear. What does it mean? In a sequence transition matrix from which they derive, and their intrinsic A➔B➔C➔D[…]➔W➔X➔Y➔Z, where each letter represents a statistical connotations make it possible to establish an important behavioral event (from a conceptual view, it is more correct to refer and useful “boundary” (that is, the significance level) to the to units of behavior occurring over time in terms of “events”), various transitions that are taken into account or shown. For statistically significant relationships may not occur between events example, regardless of the illustrative approach used, one may in direct succession (that is, between B and A, C and B etc.); therefore choose to present only those transitions that are actually, the opposite is almost always true and significantly statistically significant. This important aspect has been stressed by related events are often not in direct succession (e.g., D and A, Y Spruijt and Gispen (1984) in a pioneering study going back almost and W etc.). This makes temporal sequences difficult to four decades. In brief, this study showed the effects of ACTH on be perceived and elusive to the naked eye. Eibl-Eibesfeldt (1970) grooming behavior in rats and highlighted that observed changes himself emphasized this essential aspect by stating that should inform future interpretations of the animal’s behavior in investigations of behavior must deal with sequences that are not the specific context in which it is observed. easily perceivable and this is why researchers need improved Frontiers in Psychology 08 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 these analyses. These difficulties have been gradually reduced over the past 30 years thanks to considerable technological development and the introduction of an elegant technique suitable for studying the temporal structure of behavior: the T-pattern analysis (TPA; Magnusson, 1996, 2000, 2004, 2005, 2017, 2020; Casarrubea et al., 2015a, 2018; Magnusson et al., 2016; Casarrubea and Di Giovanni, 2020). Through this approach, recurrent sequences of events, called T-patterns, can be easily highlighted and, therefore, studied. A remarkable aspect of the method is its complete independence from the underlying time scale. Indeed, TPA allows for the study of events that occur in the order of milliseconds (e.g., those related to the firing of neurons) as well as events that span much larger time windows (e.g., those related to rodent behavior). The detection of T-patterns can be performed by means of a software tool known as THEME (Patternvision Ltd., Reykjavik, Iceland). A T-pattern can be described using the following expression: X1 ≈ dt1 X 2 ≈ dt2 X 3 … X i ≈ dti X i +1 where, X terms are the events belonging to a hypothetical T-pattern and ≈dt terms represent the time distances separating these events. Thus, the term X1 ≈ dt1 X2 indicates that the X1 event is followed dt1 time units later by the X2 event; so, Xi ≈ dti Xi + 1 symbolizes that Xi is followed dti time units later by Xi + 1. Figure 8 shows a sequence of 30 hypothetical events (letters near the axis) occurring in the context of an observation time window T0-Tx. The detection algorithm compares the distribution of each pair of events (e.g., A and B) searching for an interval in which A is followed by B more often than expected by chance. If B does not fall within such an interval, another event (e.g., C) is tested, and so on, for all the events; if event B falls within the interval, A and B represent a T-pattern encompassing only two events and will be indicated as (A B). This (A B) T-pattern is then used by the algorithm to detect higher-order patterns, e.g., [(A B) C], [(A B) (C D)] etc. Such a bottom-up detection process runs up to any level and stops when no more T-patterns are identified. The use of TPA allows for the identification of three important qualitative aspects related to the temporal structure of behavior: its variability, complexity and recursiveness (Casarrubea et al., 2021b). The variability is represented by the amount of T-patterns of different FIGURE 7 composition detected, the complexity is the length of T-patterns Hypothetical histogram of adjusted residuals representing the (i.e., the number of events in their sequences), and the differences between observed and expected values. Heading of each panel: behavioral unit antecedent to the ones indicated at recursiveness is the number of times each T-pattern is repeated. the bottom. Dark bars, positive residuals (i.e., transitions More details concerning concepts, theories and procedures occurring more often than expected); light grey bars, negative underlying the detection and analysis of T-patterns can be found residuals (i.e., transitions occurring less often than expected). See text for details. in numerous papers of Magnusson (Magnusson, 1996, 2000, 2004, 2005, 2017, 2020; Magnusson et al., 2016) and in a number of works from our laboratory and field studies (Casarrubea et al., methods of detection and analysis of the temporal dimension of 2015a, 2018; Casarrubea and Di Giovanni, 2020; Cenni et al., behavioral sequences. 2020; Gunst et al., 2020). These studies have successfully shown The study of the temporal structure of behavior was, for many that the temporal dimension of behavioral sequences represents a years, a difficult shore to reach mainly because of an unfortunate powerful structural variable to infer underlying behavioral and combination consisting of the lack of adequate computational psychological mechanisms, and explain the actions being resources compounded by the lack of efficient models to conduct performed in terms of functional components. Frontiers in Psychology 09 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 FIGURE 8 Hypothetical observation period (T0–Tx) consisting of 30 events (black letters near time axis). The [(A B) C] T-pattern becomes evident when all the remaining behavioral occurrences are left out (light grey letters). See text for details. From rodents to non-human the same vein, let us just mention a study by Lino-de-Oliveira et al. primates (2005) pertaining to locomotory behavior of rats subjected to the forced swimming test, an elegant study by Takahashi et al. (2010) Even though the above-outlined methods are not the only on rats’ ultrasonic vocalizations, and numerous studies by ones used to identify and describe the structural relationships Casarrubea and Colleagues focusing on the structure of between the different units of a behavioral repertoire, they are exploratory behavior, anxiety-related response, and reaction to fruitful approaches. After providing examples of significant pain in rats and mice, and involving various lab apparatuses such studies employing these analytical techniques across disciplines an Open Field (e.g., Casarrubea et al., 2008; Santangelo et al., and animal taxa, we discuss the benefits and challenges associated 2018), Hole Board (e.g., Casarrubea et al., 2009b, 2017) and with the use of these methods. Hot-Plate (e.g., Casarrubea et al., 2006, 2012). In non-human primates, both “time lag” and “state lag” sequential analyses have been successfully used to demonstrate Transition matrices and related analyses that female-to-male mounting in Japanese macaques is a supernormal courtship behavior that functions to focus the male Behavioral studies based on the analysis and transformation consort partner’s attention and prompt subsequent male-to- of transition matrices have produced a large number of female mounting (Gunst et al. 2022). publications straddling various scientific disciplines and including What are the strengths and weaknesses of the utilization of a wide range of animal taxa from insects (e.g., Berman et al., 2016) transition matrices? As far as benefits are concerned, whether the to reptiles (e.g., McElroy et al., 2012), fishes (e.g., Van Der Heijden applications lie in a deep understanding of the structure of et al., 1990), birds (e.g., Eens et al., 1989), rodents (e.g., Espejo and behavior in rodents or primates, these methods are unified by the Mir, 1993), all the way to non-human primates (e.g., Skorupa, possibility of presenting the results in an extremely direct and 1983) and humans (e.g., McCune and McCune, 2019). Despite the intuitive way, which makes the publications pleasant to read and common use of transition matrices, these studies employ a great the findings easy to understand for a wide and non-expert deal of different procedures, and it is beyond the scope of this audience. One only has to look at the dendrogram (Figure 3), the article to explore the enormous variability in the application of pathway diagram (Figure 4) and/or the histogram of residuals these analytical methods. That said, within the research employing (Figure 7) to immediately grasp the meaning of these graphical transition matrices the work conducted on the behavioral representations and how they relate to the original transition structure in rodents certainly stands out. This is not surprising, matrices, that would otherwise look obscure. Regarding which of since mice and rats are, by far, the preferred species in biomedical these methods may better convey information about behavioral research due to their relatively low cost, basic housing conditions, structure, probabilistic analysis and pathway diagrams make it and availability of many genetic variants that are extremely useful more intuitive than the clustering procedure and/or dendrograms. from a translational point of view (Bryda, 2013). The approaches However, both pathway diagrams and dendrograms require used in most of the articles in which transition matrices are used statistical support. This can be done by analyzing the individual in the study of rodent behavior fall into one or more of the transitions in the source matrices and adding the corresponding methods outlined above. adjusted residuals that stand for the different transitions in Among these studies, pioneering works include those by statistical terms. Spruijt and Colleagues, in which adjusted residuals are Let us now address some problems associated with using predominantly employed (Spruijt and Gispen, 1984; Spruijt, 1992; transition matrices. Unlike the simplistic graphical representations Vanderschuren et al., 1996; Van Den Berg et al., 1999) and those featured in Figures 3, 4, 7 that contain only three behavioral units, by Espejo and Colleagues, that often involves the simultaneous use a typical transition matrix is a table containing dozens or even of probabilistic evaluations and aggregative clustering techniques hundreds of numbers. Therefore, reading, analyzing, and (Espejo and Mir, 1993, 1994; Espejo et al., 1994; Espejo, 1997). In interpreting a transition matrix is generally no simple matter not Frontiers in Psychology 10 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 even an option for researchers accustomed to generating matrix that the temporal structure of stone play sequences with genital analysis from data sets. Beyond the considerable work required to stone-tapping and genital stone-rubbing performed by males was obtain the event log-files from which a transition matrix is derived less structurally flexible and less exaggerated than that of stone (see section “Physiology of Behavior”), the matrix should then play sequences without genital stone-tapping and genital stone- be subjected to specific transformations that direct the subsequent rubbing, suggesting functional attributes of these two specific analysis toward clustering, probabilities, and/or residuals. Once behavioral units (Cenni et al., 2020). Thus, the performance of this step is completed, the matrix is still an analytical item that genital stone-tapping and genital stone-rubbing by male Balinese only contains a large amount of numerical data and requires long-tailed macaques could be an example of stone play actions graphical treatment to be usable for interpretation (i.e., to obtain being functionally recycled into stone tool-assisted masturbation dendrograms, probability pathway diagrams, and histograms with (Cenni et al., 2020). This result supports the view that object play residuals). In other words, one of the main problems associated can serve as a pool of behavioral variability and has the exaptive with the use of transition matrices is the enormous amount of potential to be subsequently co-opted into tool use (Leca and time required for their processing, analysis, and graphical Gunst, in press). Second, in a free-ranging population of Japanese transformation to make them intelligible. Yet, there is an even macaques in which female-to-male mounting was identified as a greater problem. The dendrograms, pathway diagrams and culturally-maintained form of non-conceptive sex, TPA showed histograms containing positive and negative residuals are that the occurrence of female-to-male mounting conferred further analogous to photographs with a very long exposure: on the one functional constraints to mating sequences with more hand, they describe the entire observation period but, on the other hierarchically organized and less repeatable courtship behaviors hand, they lack one essential feature for any behavior: its temporal than in mating sequences without female-to-male mounting dimension. As pointed out by Eibl-Eibesfeldt (1970), behavior is (Gunst et al., 2020). This result supports the view that female-to- structured on the basis of a flow of events that runs through time. male mounting in Japanese macaques is a supernormal courtship display more efficient than species-typical female-to-male sexual solicitations at prompting subsequent male-to-female mounts Temporal dynamics of behavior (Gunst et al. 2022). In humans, TPA has been widely applied to detect tactical Over the past three decades or so, the search and analysis of movements or prevent injuries in several sports and physical T-patterns have been successfully and widely used by scientists activities, such as boxing (Pic and Jonsson, 2021), taekwondo from a number of different fields, including psychology, (Gutiérrez-Santiago et al., 2020), and football (Prieto-Lage et al., psychiatry, computer science, physiology, biology, and sport 2020), as well as to explore the relationships between impulsivity science. The reasons underlying the inter-disciplinary popularity and physical activity (Castañer et al., 2020). TPA was also used in of TPA is that this technique offers researchers a great level of the study of human-animal interactions (Kerepesi et al., 2005), detail in the temporal structure of the behaviors under study. TPA human-robot interactions (Kerepesi et al., 2006), hormones and has been successfully employed to explore the temporal behavior (Hirschenhauser et al., 2002), decision-making processes organization of behavior in numerous species, ranging from (Pic et al., 2021), eye-blinking behavior (Brill and Schwab, 2020), rodents to primates. movement and behavioral disorders (Aiello et al., 2020), In rodents, TPA allowed researchers to describe the qualitative neuropsychiatric diseases (Lyon and Kemp, 2004; Kemp et al., and quantitative aspects of feeding behavior in rats under different 2008, 2016; Sandman et al., 2012) and the behavior of dietary regimes (Casarrubea et al., 2019a), analyze anxiety-related preschool-age children (Santoyo et al., 2020). behaviors in different rat strains (Casarrubea et al., 2010, 2011b, The first advantage of using TPA in behavioral research is 2013a,b, 2014, 2015b, 2016, 2020a, 2021a), examine directly related to the outputs of this approach: providing detailed psychostimulant-evoked route-tracing stereotypies in mice information, otherwise impossible to access, about the temporal (Bonasera et al., 2008), as well as generate a model of Tourette’s structure of behavior. To the best of our knowledge, no other syndrome (Santangelo et al., 2018) and a model of Parkinson’s analytical techniques allow for the study of the temporal dynamics Disease (Casarrubea et al., 2019b) in rats. of behavior at such a fine-grained resolution. Another benefit of In non-human primates, TPA has recently been used to TPA is its independence from the observed time window; whether demonstrate that the temporal dynamics of a behavior is it is a few seconds (typical of neuron firing measurements), a few informative to infer its function, and this, in two different minutes (typical of experimentally-induced behavioral behavioral domains: tool use (Cenni et al., 2020) and measurements in lab rodents), or hours (typical of the assessment non-conceptive sex (Gunst et al., 2020). First, in a free-ranging of a daily activity budget in free-ranging primates), the principles population of Balinese long-tailed macaques in which versatile of T-pattern detection remain the same. Finally, TPA allows stone play was identified as a behavioral tradition, the monkeys researchers to obtain three essential qualitative characteristics of were reported to repeatedly tap and rub stones onto their genital the temporal texture of behavior, namely its variability, complexity, area (Pelletier et al., 2017). Despite those two stone-directed and recursiveness (Casarrubea et al., 2019a, 2021b). The actions being integrated into stone play episodes, TPA revealed evaluation of such qualitative aspects, combined with the Frontiers in Psychology 11 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 possibility to analyze specific events in the structure of the we sought to provide some theoretical background for beginners identified sequences allow for a further understanding of the in the exploration of behavioral structure and for already behavior. For example, comparisons between rats treated with experienced researchers who wish to implement some of the standard diet and rats treated with hyperglycidic diet revealed approaches outlined here. Structural behavioral analyses provide profound variations in those qualitative parameters. Specifically, substantial benefits that, by far, compensate for the large amount animals under standard diet showed a behavior characterized by of time required for their use. The main advantage of these fewer T-patterns of different composition (i.e., lower temporal methods consists in producing qualitative and quantitative variability) which are more often repeated (i.e., higher descriptions of complex behavioral dynamics that would recursiveness); on the other hand, rats under a hyperglycidic diet otherwise be difficult or impossible to obtain. showed a behavior characterized by a noticeably higher number of different T-patterns (i.e., higher temporal variability) but that are repeated less often (i.e., lower recursiveness). The evaluation Author contributions of the qualitative aspects of these patterns showed a significantly higher percentage of T-patterns specifically associated with two MC conceived the study. J-BL, NG, GJ, GC, and MC wrote the behavioral units, the so called Focusing-Sniffing (i.e., when the rat manuscript. MC, J-BL, NG, and SA conducted field-and sniffs the rim of the pellet box without inserting the head inside) laboratory-based research and performed data analysis. MC and Feeding (i.e., when the rat inserts the head inside the pellet prepared tables and Figures. GC, GG, J-BL, NG, GJ, and MP box and eats) in the hyperglicidic group than in the control group. reviewed and edited the manuscript. All authors contributed to These results allowed researchers to hypothesize an increased the article and approved the submitted version. salience of food-related stimuli in rats under hyperglycidic diet and a behavior highly evocative of craving (Casarrubea et al., 2019a). Funding One of the main challenges associated with the use of TPA is to fine-tune the use of the software with the data at hand. The most J-BL’s research was funded by Natural Sciences and recent versions of the software tool utilized to perform T-patterns’ Engineering Research Council of Canada (NSERC, Discovery detection analysis do feature numerous options; a thorough Grant #: 2015-06034 to J-BL). MC, SA, and GC’s research was knowledge and practice of each of them is crucial to the reliability funded by a grant from the University of Palermo, Italy. and validity of any outputs. Conversely, poor use of the software parameters can lead to the detections of T-patterns that might not be salient to the research question or difficult to interpret. An Conflict of interest additional set of difficulties pertaining to the use of TPA is common with the analysis of transition matrices. TPA software also require The authors declare that the research was conducted in the event log-files to perform T-pattern detection. The creation and absence of any commercial or financial relationships that could organization of such log-files before import into the software tool, be construed as a potential conflict of interest. albeit not complex in theory, are extremely time-consuming tasks The reviewer MB declared a past co-authorship with the (see section Physiology of behavior.). Although we often speak of authors MC and GG to the handling Editor. “T-pattern analysis” to refer to the entire analytical process, this essential preparatory step should be kept in mind. Publisher’s note Conclusion All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated In this paper, we aimed to present the main methods that have organizations, or those of the publisher, the editors and the accompanied us in the structural analysis of behavior over the past reviewers. Any product that may be evaluated in this article, or two decades. Our goal was not to present an excursus on the use claim that may be made by its manufacturer, is not guaranteed or of multivariate analyses in this line of behavioral research. Instead, endorsed by the publisher. References Aiello, S., Crescimanno, G., Di Giovanni, G., and Casarrubea, Ashby, W. R. (1962). Principles of the self-organizing system. In: FoersterH Von M. (2020). T-patterns in the study of movement and behavioral and ZopfGW Jr (esd.) Principles of self-organization: Transactions of the University disorders. Physiol. Behav. 215:112790. doi: 10.1016/j.physbeh.2019. of Illinois symposium. London, UK: Pergamon Press. pp. 255–278. 112790 Berman, G. J., Bialek, W., and Shaevitz, J. W. (2016). Predictability and hierarchy Anderson, P. W. (1972). More is different. Science 177, 393–396. doi: 10.1126/ in drosophila behavior. Proc. Natl. Acad. Sci. U. S. A. 113, 11943–11948. doi: science.177.4047.393 10.1073/pnas.1607601113 Frontiers in Psychology 12 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 Bonasera, S. J., Schenk, A. K., Luxenberg, E. J., and Tecott, L. H. (2008). A novel Casarrubea, M., Sorbera, F., and Crescimanno, G. (2008). Multivariate analysis of method for automatic quantification of psychostimulant-evoked route-tracing the modifications induced by an environmental acoustic cue on rat exploratory stereotypy: application to Mus musculus. Psychopharmacology 196, 591–602. doi: behavior. Physiol. Behav. 93, 687–696. doi: 10.1016/j.physbeh.2007.11.010 10.1007/s00213-007-0994-6 Casarrubea, M., Sorbera, F., and Crescimanno, G. (2009a). Structure of rat Brill, M., and Schwab, F. (2020). T-pattern analysis and spike train dissimilarity behavior in hole-board: I. multivariate analysis of response to anxiety. Physiol. for the analysis of structure in blinking behavior. Physiol. Behav. 227:113163. doi: Behav. 96, 174–179. doi: 10.1016/j.physbeh.2008.09.025 10.1016/j.physbeh.2020.113163 Casarrubea, M., Sorbera, F., and Crescimanno, G. (2009b). Structure of rat Bryda, E. C. (2013). The mighty mouse: the impact of rodents on advances in behavior in hole-board: II multivariate analysis of modifications induced by biomedical research. Mo. Med. 110, 207–211. PMID: 23829104 diazepam. Physiol. Behav. 96, 683–692. doi: 10.1016/j.physbeh.2009.01.005 Casarrubea, M., Aiello, S., Di Giovanni, G., Santangelo, A., Palacino, M., and Casarrubea, M., Sorbera, F., Magnusson, M., and Crescimanno, G. (2010). Crescimanno, G. (2019a). Combining quantitative and qualitative data in the study Temporal patterns analysis of rat behavior in hole-board. Behav. Brain Res. 208, of feeding behavior in male Wistar rats. Front. Psychol. 10:881. doi: 10.3389/ 124–131. doi: 10.1016/j.bbr.2009.11.013 fpsyg.2019.00881 Casarrubea, M., Sorbera, F., Magnusson, M. S., and Crescimanno, G. (2011b). Casarrubea, M., Davies, C., Faulisi, F., Pierucci, M., Colangeli, R., Partridge, L., T-pattern analysis of diazepam-induced modifications on the temporal organization et al. (2015b). Acute nicotine induces anxiety and disrupts temporal pattern of rat behavioral response to anxiety in hole board. Psychopharmacology 215, organization of rat exploratory behavior in hole-board: a potential role for the lateral 177–189. doi: 10.1007/s00213-010-2123-1 habenula. Front. Cell. Neurosci. 9:197. doi: 10.3389/fncel.2015.00197 Casarrubea, M., Sorbera, F., Santangelo, A., and Crescimanno, G. (2011a). Casarrubea, M., Davies, C., Pierucci, M., Colangeli, R., Deidda, G., Santangelo, A., Learning influence on the behavioral structure of rat response to pain in hot-plate. et al. (2021a). The impact of chronic daily nicotine exposure and its overnight Behav. Brain Res. 225, 177–183. doi: 10.1016/j.bbr.2011.07.020 withdrawal on the structure of anxiety-related behaviors in rats: role of the lateral Casarrubea, M., Sorbera, F., Santangelo, A., and Crescimanno, G. (2012). The Habenula. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 105:110131. doi: 10.1016/j. effects of diazepam on the behavioral structure of the rat's response to pain in the pnpbp.2020.110131 hot-plate test: anxiolysis vs. pain modulation. Neuropharmacology 63, 310–321. doi: Casarrubea, M., and Di Giovanni, G. (2020). Application of T-pattern analysis in 10.1016/j.neuropharm.2012.03.026 the study of the organization of behavior. Physiol. Behav. 227:113138. doi: 10.1016/j. Castañer, M., Aiello, S., Prat, Q., Andueza, J., Crescimanno, G., and Camerino, O. physbeh.2020.113138 (2020). Impulsivity and physical activity: a T-pattern detection of motor behavior Casarrubea, M., Di Giovanni, G., Crescimanno, G., Rosa, I., Aiello, S., Di profiles. Physiol. Behav. 219:112849. doi: 10.1016/j.physbeh.2020.112849 Censo, D., et al. (2019b). Effects of substantia Nigra pars compacta lesion on the Cenni, C., Casarrubea, M., Gunst, N., Vasey, P., Pellis, S., Wandia, I. N., et al. behavioral sequencing in the 6-OHDA model of Parkinson’s disease. Behav. Brain (2020). Inferring functional patterns of tool use behavior from the temporal Res. 362, 28–35. doi: 10.1016/j.bbr.2019.01.004 structure of object play sequences in a non-human primate species. Physiol. Behav. Casarrubea, M., Faulisi, F., Caternicchia, F., Santangelo, A., Di Giovanni, G., 222:112938. doi: 10.1016/j.physbeh.2020.112938 Benigno, A., et al. (2016). Temporal patterns of rat behaviour in the central platform Cenni, C., Christie, J. B. A., Van der Pant, Y., Gunst, N., Vasey, P. L., Wandia, I. N., of the elevated plus maze. Comparative analysis between male subjects of strains et al. (2022). Do monkeys use sex toys? Evidence of stone tool-assisted masturbation with different basal levels of emotionality. J. Neurosci. Methods 268, 155–162. doi: in free-ranging long-tailed macaques. Ethology 128, 632–646. 10.1016/j.jneumeth.2015.07.024 Eens, M., Pinxten, R., and Verheyen, R. F. (1989). Temporal and sequential Casarrubea, M., Faulisi, F., Pensabene, M., Mendola, C., Dell'Utri, R., Cardaci, M., organisation of song bouts in the starling. Ardea 77, 75–86. et al. (2017). Effects of the benzodiazepine inverse agonist FG7142 on the structure of anxiety-related behavior of male Wistar rats tested in hole board. Eibl-Eibesfeldt, I. (1970). Ethology: The biology of behavior. Holt, Rinehart and Psychopharmacology 234, 381–391. doi: 10.1007/s00213-016-4474-8 Winston: New York, NY. Casarrubea, M., Faulisi, F., Raso, G., Aiello, S., and Crescimanno, G. (2020b). Espejo, E. F. (1997). Structure of the mouse behaviour on the elevated plus-maze Early alterations of the behavioural structure of mice affected by Duchenne test of anxiety. Behav. Brain Res. 86, 105–112. doi: 10.1016/S0166-4328(96)02245-0 muscular dystrophy and tested in open-field. Behav. Brain Res. 386:112609. doi: Espejo, E. F., and Mir, D. (1993). Structure of the rat’s behaviour in the hot plate 10.1016/j.bbr.2020.112609 test. Behav. Brain Res. 56, 171–176. doi: 10.1016/0166-4328(93)90035-O Casarrubea, M., Faulisi, F., Sorbera, F., and Crescimanno, G. (2015c). The effects Espejo, E. F., and Mir, D. (1994). Differential effects of weekly and daily exposure of different basal levels of anxiety on the behavioral shift analyzed in the central to the hot plate on the rat’s behavior. Physiol. Behav. 55, 1157–1162. doi: platform of the elevated plus maze. Behav. Brain Res. 281, 55–61. doi: 10.1016/j. 10.1016/0031-9384(94)90404-9 bbr.2014.12.021 Espejo, E. F., Stinus, L., Cador, M., and Mir, D. (1994). Effects of morphine and Casarrubea, M., Jonsson, G. K., Faulisi, F., Sorbera, F., Di Giovanni, G., naloxone on behaviour in the hot plate test: an ethopharmacological study in the Benigno, A., et al. (2015a). T-pattern analysis for the study of temporal structure of rat. Psychopharmacology 113, 500–510. doi: 10.1007/BF02245230 animal and human behavior: a comprehensive review. J. Neurosci. Methods 239, 34–46. doi: 10.1016/j.jneumeth.2014.09.024 Everitt, B. S. (1977). The analysis of contingency tables. Wiley, New York, NY. Casarrubea, M., Magnusson, M. S., Anguera, M. T., Jonsson, G. K., Castañer, M., Faraone, S. V., and Dorfman, D. D. (1987). Lag sequential analysis: robust Santangelo, A., et al. (2018). T-pattern detection and analysis for the discovery of statistical methods. Psychol. Bull. 101, 312–323. doi: 10.1037/0033-2909.101.2.312 hidden features of behaviour. J. Neurosci. Methods 310, 24–32. doi: 10.1016/j. Gunst, N., Casarrubea, M., Vasey, P., and Leca, J.-B. (2020). Is female-male jneumeth.2018.06.013 mounting functional? An analysis of the temporal patterns of sexual behaviors in Casarrubea, M., Magnusson, M. S., Roy, V., Arabo, A., Sorbera, F., Santangelo, A., Japanese macaques. Physiol. Behav. 223:112983. doi: 10.1016/j.physbeh.2020.112983 et al. (2014). Multivariate temporal pattern analysis applied to the study of rat Gunst, N., Leca, J.-B., and Vasey, P. L. (2022a). Sexual adaptation: is female-male behavior in the elevated plus maze: methodological and conceptual highlights. J. mounting a supernormal courtship display in Japanese macaques? Behaviour 159, Neurosci. Methods 234, 116–126. doi: 10.1016/j.jneumeth.2014.06.009 999–1027. Casarrubea, M., Palacino, M., Brancato, A., Lavanco, G., Cannizzaro, C., and Gunst, N., Leca, J.-B., and Vasey, P. L. (2022b). When males have females on their Crescimanno, G. (2021b). Detection of a temporal structure in the rat behavioural backs: Male’s tolerance, solicitation, and use of female-male mounting in Japanese response to an aversive stimulation in the emotional object recognition (EOR) task. macaques. Am. J. Primatol. 84:e23395. doi: 10.1002/ajp.23395 Physiol. Behav. 238:113481. doi: 10.1016/j.physbeh.2021.113481 Gunst, N., Pellis, S. M., Wandia, I. N., and Leca, J.-B. (2022). A tool to act blind? Casarrubea, M., Pierucci, M., Aiello, S., Cassar, D., Deidda, G., Crescimanno, G., Object-assisted eye-covering as a self-handicapping behavior and social play signal et al. (2020a). Effects of chronic nicotine on the temporal structure of anxiety- in Balinese long-tailed macaques. Anim. Cogn. doi: 10.1007/s10071-022-01707-4 related behavior in rats tested in hole-board. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 96:109731. doi: 10.1016/j.pnpbp.2019.109731 Gutiérrez-Santiago, A., Pereira-Rodríguez, R., and Prieto-Lage, I. (2020). Detection of the technical and tactical motion of the scorable movements in Casarrubea, M., Roy, V., Sorbera, F., Magnusson, M. S., Santangelo, A., Arabo, A., taekwondo. Physiol. Behav. 217:112813. doi: 10.1016/j.physbeh.2020.112813 et al. (2013a). Significant divergences between the temporal structure of the behavior in Wistar and in the spontaneously more anxious DA/Han strain of rats tested in Haberman, S. J. (1973). The analysis of residuals in cross-classified tables. elevated plus maze. Behav. Brain Res. 250, 166–173. doi: 10.1016/j.bbr.2013.05.016 Biometrics 29, 205–220. doi: 10.2307/2529686 Casarrubea, M., Roy, V., Sorbera, F., Magnusson, M. S., Santangelo, A., Arabo, A., Hemerik, L., Bukovinszky, T., Gols, R., van Lenteren, J. C., and Vet, L. E. M. et al. (2013b). Temporal structure of the rat’s behavior in elevated plus maze test. (2006). Enter the matrix: how to analyze the structure of behavior. Behav. Res. Behav. Brain Res. 237, 290–299. doi: 10.1016/j.bbr.2012.09.049 Methods 38, 357–363. doi: 10.3758/BF03192788 Casarrubea, M., Sorbera, F., and Crescimanno, G. (2006). Effects of 7-OH-DPAT Hirschenhauser, K., Frigerio, D., Grammer, K., and Magnusson, M. S. (2002). and U 99194 on the behavioral response to hot plate test, in rats. Physiol. Behav. 89, Monthly patterns of testosterone and behavior in prospective fathers. Horm. Behav. 552–562. doi: 10.1016/j.physbeh.2006.07.014 42, 172–181. doi: 10.1006/hbeh.2002.1815 Frontiers in Psychology 13 frontiersin.org Casarrubea et al. 10.3389/fpsyg.2022.1033561 Hughes, M. M. (1978). Sequential analysis of exploration and play. Int. J. Behav. ecological perspectives. eds. M. Bekoff and J. A. Byers (Cambridge, UK: Cambridge Dev. 1, 83–97. doi: 10.1177/016502547800100106 University Press), 115–140. Hughes, M. M. (1979). Exploration and play re-visited: a hierarchical analysis. Int. Pellis, S. M., Pellis, V., Pelletier, A., and Leca, J.-B. (2019). Is play a behavior J. Behav. Dev. 2, 215–224. doi: 10.1177/016502547900200301 system, and, if so, what kind? Behav. Proc. 160, 1–9. doi: 10.1016/j. beproc.2018.12.011 Kemp, A. S., Fillmore, P. T., Lenjavi, M. R., Lyon, M., Chicz-DeMet, A., Touchette, P. E., et al. (2008). Temporal patterns of self-injurious behavior correlate Pic, M., and Jonsson, G. K. (2021). Professional boxing analysis with T-patterns. with stress hormone levels in the developmentally disabled. Psychiatry Res. 157, Physiol. Behav. 232:113329. doi: 10.1016/j.physbeh.2021.113329 181–189. doi: 10.1016/j.psychres.2007.04.003 Pic, M., Navarro-Adelantado, V., and Jonsson, G. K. (2021). Exploring playful Kemp, A. S., Lenjavi, M. R., Touchette, P. E., Pincus, D., Magnusson, M. S., and asymmetries for gender-related decision-making through T-pattern analysis. Sandman, C. A. (2016). “The self-Organization of Self-Injurious Behavior as Physiol. Behav. 236:113421. doi: 10.1016/j.physbeh.2021.113421 revealed through temporal pattern analyses,” in Discovering hidden temporal Prieto-Lage, I., Louzao-Neira, I., Argibay-González, J. C., and patterns in behavior and interaction, vol. 111 eds. M. S. Magnusson, J. K. Burgoon Gutiérrez-Santiago, A. (2020). Injury patterns of professional footballers in the and M. Casarrubea (New York, NY, USA: Springer), 101–124. Spanish first division during the 2017-2018 seasons. Physiol. Behav. 224:113052. doi: Kerepesi, A., Jonsson, G. K., Emiklósi, Á., Topál, J., Csányi, V., and 10.1016/j.physbeh.2020.113052 Magnusson, M. S. (2005). Detection of temporal patterns in dog-human interaction. Quera, V., and Estany, E. (1984). ANSEC: a BASIC package for lag sequential Behav. Process. 70, 69–79. doi: 10.1016/j.beproc.2005.04.006 analysis of observational data. Behav. Res. Methods Instrum. Comput. 16, 303–306. Kerepesi, A., Kubinyi, E., Jonsson, G. K., Magnusson, M. S., and Emiklósi, A. doi: 10.3758/BF03202408 (2006). Behavioural comparison of human-animal (dog) and human-robot (AIBO) Rasa, O. A. E. (1984). A motivational analysis of object play in juvenile dwarf interactions. Behav. Process. 73, 92–99. doi: 10.1016/j.beproc.2006.04.001 mongooses (Helogale undulata rufula). Anim. Behav. 32, 579–589. doi: 10.1016/ Leca, J. B., and Gunst, N. (in press). The exaptive potential of (object) play S0003-3472(84)80295-X behavior. Int. J. Play. Sackett, G. P. (1979). “The lag sequential analysis of contingency and cyclicity in Lino-de-Oliveira, C., De Lima, T. C., and de Pádua Carobrez, A. (2005). Structure behavioral interaction research,” in Handbook of infant development. ed. J. D. of the rat behaviour in the forced swimming test. Behav. Brain Res. 158, 243–250. Osofsky (New York, NY: Wiley), 623–649. doi: 10.1016/j.bbr.2004.09.004 Sandman, C. A., Kemp, A. S., Mabini, C., Pincus, D., and Magnusson, M. (2012). Lyon, M., and Kemp, A. S. (2004). Increased temporal patterns in choice The role of self-injury in the organisation of behaviour. J. Intellect. Disabil. Res. 56, responding and altered cognitive processes in schizophrenia and mania. 516–526. doi: 10.1111/j.1365-2788.2012.01552.x Psychophamacology (Berl.) 172, 211–219. doi: 10.1007/s00213-003-1646-0 Santangelo, A., Bortolato, M., Mosher, L. J., Crescimanno, G., Di Giovanni, G., Magnusson, M. S. (1996). Hidden real-time patterns in intra-and inter-individual Cassioli, E., et al. (2018). Behavioral fragmentation in the D1CT-7 mouse model of behavior: description and detection. Eur. J. Psychol. Assess. 12, 112–123. doi: Tourette's syndrome. CNS Neurosci. Ther. 24, 703–711. doi: 10.1111/cns.12789 10.1027/1015-5759.12.2.112 Santoyo, C., Jonsson, G. K., Anguera, M. T., Portell, M., Allegro, A., Magnusson, M. S. (2000). Discovering hidden time patterns in behavior: Colmenares, L., et al. (2020). T-patterns integration strategy in a longitudinal study: T-patterns and their detection. Behav. Res. Methods Instrum. Comput. 32, 93–110. a multiple case analysis. Physiol. Behav. 222:112904. doi: 10.1016/j. doi: 10.3758/BF03200792 physbeh.2020.112904 Magnusson, M. S. (2004) in Repeated patterns in behavior and other biological Short, R., and Horn, J. (1984). Some notes on factor analysis of behavioural data. phenomena. eds. D. K. Oller and U. Griebel (Cambridge, MA, USA: The MIT Press) Behaviour 90, 203–214. doi: 10.1163/156853984X00623 Magnusson, M. S. (2005). “Understanding social interaction: discovering hidden Skorupa, J. P. (1983). Monkeys and matrices: a second look. Oecologia (Berlin). 57, structure with model and algorithms,” in The hidden structure of interaction: From 391–396. doi: 10.1007/BF00377185 neurons to culture patterns, eds. G. Riva, L. Anolli, S. Duncan Jr., M. S. Magnusson Spruijt, B. M. (1992). Progressive decline in social attention in aging rats: an (Amsterdam, The Netherlands: IOS Press), 3–22. information-statistical method. Neurobiol. Aging 13, 145–151. doi: Magnusson, M. S. (2017). Why search for hidden repeated temporal behavior 10.1016/0197-4580(92)90022-P patterns: T-pattern analysis with Theme. Int. J. Clin. Pharmacol. Pharmacother. Spruijt, B. M., and Gispen, W. H. (1984). Behavioral sequences as an easily 2:128. doi: 10.15344/2456-3501/2017/128 quantifiable parameter in experimental studies. Physiol. Behav. 32, 707–710. doi: Magnusson, M. S. (2020). T-patterns, external memory and mass-societies in 10.1016/0031-9384(84)90182-3 proteins and humans: in an eye-blink the naked ape became a string-controlled Takahashi, N., Kashino, M., and Hironaka, N. (2010). Structure of rat ultrasonic citizen. Physiol. Behav. 227:113146. doi: 10.1016/j.physbeh.2020.113146 vocalizations and its relevance to behavior. PLoS One 5:e14115. doi: 10.1371/journal. Magnusson, M. S., Burgoon, J. K., and Casarrubea, M. (2016). Discovering hidden pone.0014115 temporal patterns in behavior and interaction, Springer: New York, NY, USA. Tinbergen, N. (1951a). “Chapter II: behavior as a reaction to external stimuli,” in Martin, P., and Caro, T. M. (1985). On the functions of play and its role in The study of instinct (UK: Clarendon Press Publication). behavioral development. Adv. Study Behav. 15, 59–103. doi: 10.1016/ Tinbergen, N. (1951b). “Chapter III: the internal factors responsible for the S0065-3454(08)60487-8 spontaneity of behaviour,” in The study of instinct (UK: Clarendon Press Publication). McCune, D., and McCune, L. (2019). Counting your chickens with Markov Tinbergen, N. (1963). On aims and methods of ethology. Z. Tierpsychol. 20, chains. Math. Mag. 92, 163–172. doi: 10.1080/0025570X.2019.1561097 410–433. doi: 10.1111/j.1439-0310.1963.tb01161.x McElroy, E. J., McBrayer, L. D., Williams, S. C., Anderson, R. A., and Reilly, S. M. Van Den Berg, C. L., Van Ree, J. M., and Spruijt, B. M. (1999). Sequential analysis (2012). Sequential analyses of foraging behavior and attack speed in ambush and of juvenile isolation-induced decreased social behavior in the adult rat. Physiol. widely foraging lizards. Adapt. Behav. 20, 16–31. doi: 10.1177/1059712311426800 Behav. 67, 483–488. Moran, G. (1985). Behavioral description and its impact on functional inference. Van Der Heijden, P. G. M., De Vries, H., and Van Hooff, J. (1990). Correspondence Behav. Brain Sci. 8, 186–187. doi: 10.1017/S0140525X0002032X analysis of transition matrices, with special attention to missing entries and Mos, J., Olivier, B., and van der Poel, A. M. (1987). Modulatory actions of asymmetry. Anim. Behav. 40, 49–64. doi: 10.1016/S0003-3472(05)80665-7 benzodiazepine receptor ligands on agonistic behaviour. Physiol. Behav. 41, 265–278. van Lier, H., Coenen, A. M. L., and Drinkenburg, W. H. I. M. (2003). Behavioral doi: 10.1016/0031-9384(87)90363-5 transitions modulate hippocampal electroencephalogram correlates of open field Pelletier, A. N., Kaufmann, T., Mohak, S., Milan, R., Nahallage, C. A. D., Huffman, M. A., behavior in the rat: support for a sensorimotor function of hippocampal rhythmical et al. (2017). Behavior systems approach to object play: stone handling repertoire as a synchronous activity. J. Neurosci. 23, 2459–2465. doi: 10.1523/ measure of propensity for complex foraging and percussive tool use in the genus Macaca. JNEUROSCI.23-06-02459.2003 Anim. Behav. Cogn. 4, 455–473. doi: 10.26451/abc.04.04.05.201 Vanderschuren, L. J. M. J., Spruijt, B. M., Hol, T., Niesink, R. J. M., and Van Pellis, S. M., and Pellis, V. C. (1998). “The structure-function interface in the Ree, J. M. (1996). Sequential analysis of social play behavior in juvenile rats: effects analysis of play FIGUREhting,” in Animal play: Evolutionary, comparative and of morphine. Behav. Brain Res. 72, 89–95. doi: 10.1016/0166-4328(96)00060-5 Frontiers in Psychology 14 frontiersin.org