If scientists have been able to track individuals with acoustic pingers, tracking a whole school has always been a challenge. Since 1996, we have been using a low-frequency omnidirectional multi-beam sonar to examine the behaviour of schools of different species of small pelagics. Our methodology was to produce very few perturbations during our observations in order to record unbiased movements of schools. Our research vessel was drifting in the middle of clusters of fish schools, with all the power block and pumps off, and the light shut during night records. A correct sampling method and a simple standardization of omnidirectional sonar, allowed us to observe fish schools far from the vessel (>400 meters) through a high sampling volume.
The swimming pattern of each school was recorded in two dimensions (x, y) + time (t), allowing estimates of basic movement parameters (directions and speeds). We have obtained information on schooling behaviour (attraction/fusion/fission) as well as spatial dynamics of schools within clusters. Observations during day-night transition periods gave some new insights into schooling behaviour, e.g. some unexpected high aggregative behaviour observed during some nights. We show and discuss a specific index of horizontal movement of fish schools that can be defined from the measured parameters, and its possible use to classify movements of school.

Ultrasonic telemetry can provide very precise data on the vertical and horizontal movements of pelagic fishes, but usually for only relatively short durations (1-3 days). Transmitters can be constructed with battery lives of one to several weeks, and in several studies tagged tunas have been relocated days to weeks after the initial track. Relocation of tagged fish is, however, generally not attempted. Efforts to track several tunas tagged with ultrasonic transmitters simultaneously have generally not been successful, as school fidelity appeared to be short lived. Strong correlations have been found between vertical movements of tunas monitored by ultrasonic tags and assumed prey species when vertical movements of the latter were recorded with sophisticated sonar systems. In contrast, archival (i.e. electronic data recording tags) have been successfully used to follow the movements of pelagic fishes from months to years. Although archival tags provide very precise data on vertical movements, light-based geolocations are not sufficiently accurate nor precise to allow inferences of schooling behaviors, or fish aggregation to specific oceanographic or geographic features. Long term recording of vertical movements of several individual bigeye tuna equipped with archival tags at Cross Seamount (southwest of the Island of Hawaii) showed long term synchronicity of extensive vertical movements, clearly implying that these individuals remained in the same school. Moreover, vertical movement data showed behaviors presumed to be characteristic of fishes associated with seamounts for up to four weeks, strongly suggesting that the fish remained at the seamount for this period. This conclusion, however, could not be substantiated due to the error terms associated with light-based geolocation data previously described. I suggest that advancements in understanding schooling and aggregation of pelagic fishes are more likely to come through use of ultrasonic telemetry, rather than with archival tags. And for this field to truly advance, the following improvements in tracking techniques are recommended.
1. Use a dedicated tracking vessel capable of searching for fish equipped with ultrasonic transmitters after the initial tracking period is completed. If the fish can be relocated, long term movement patterns and associations with specific oceanographic and geographic features can be inferred. This will, however, require flexible ship scheduling which has heretofore generally not been possible because of high operational expenses and/or high demand for ship time.
2. Make simultaneous observations of oceanographic conditions around and away from fish being tracked. This will require the use of a tracking vessel and a separate oceanographic vessel. A small highly maneuverable tracking vessel is clearly preferable, while a large oceanographic ship is best suited to mapping currents, depth-temperature profiles, fronts, bottom topography (if appropriate), and forage densities. The larger oceanographic vessel could also serve as a floating "hotel", allowing the crew aboard the tracking vessel to be relieved at regular intervals. With this system, the duration of the tracks of individual fish could be significantly extended. Moreover, if the oceanographic vessel is large enough, and the tracking vessel small enough, the former could be used to transport the latter to tracking sites considerable distances from port.
3. Make better use of advanced sonar capabilities to monitor simultaneous movements of individual fish carrying ultrasonic transmitter, other members of its school, and prey species.

Awareness of the potential effects of military and industrial sound on protected marine wildlife has heightened over the last decade. In one recent case, naval sonars in Bahamian waters appear to have caused the strandings of seventeen protected cetaceans; seven subsequently died. However, despite the acute need to understand how marine animals use and respond to sound, we have learned relatively little. Increased understanding requires precise measurements of acoustic stimuli and associated responses, data that are difficult to obtain for captives and nearly impossible to obtain for submerged, free-ranging individuals. A promising solution involves attachment of acoustic recording "tags" to individual animals to measure sound exposure and associated response as accurately as possible. Early experiments with these tags yielded surprising results, showing that the acoustic recorders could monitor not only exposure but behavioral and physiological response indicators as well. These results encouraged the development of a miniature, general-purpose acoustic recorder for use with a variety of protected species. Prototypes developed by Greeneridge Sciences sample acoustics with 16-bit resolution at bandwidths up to 6 kHz, as well as temperature and depth with 12-bit resolution. Acoustic bandwidths up to 14 kHz can be obtained if temperature and depth are not recorded. Constant acoustic sampling at 2 kHz fills the 288-MB solid-state flash disk in 20 hours. Low-power three-volt electronics allow a single field-replaceable lithium battery to power the entire tag. As of September 2002 the design had been deployed with success on blue whales and northern fur seals.

Many of the collective behaviours exhibited by fish schools can only be understood by considering the very large number of interactions among group members. Individual-based computer models are a very useful analytical tool to study such groups. Here I use such a modelling approach to investigate how individual behaviours result in collective patterns, including loose 'swarms', polarised groups, and the generation of a 'torus' formation where individuals perpetually rotate around an empty core. Such major group-level behavioural transitions are related to minor changes in individual-level interactions. I also present evidence for collective memory in animal groups (where the previous history of group structure influences the collective behaviour exhibited as individual interactions change) during the transition of a group from one kind of collective behaviour to another. Differences among organisms (e.g. behavioural state, age) are likely to influence the positions adopted by individuals within groups. I show how individuals employing simple, local, rules of thumb, can accurately change their spatial position within the group as a whole. This is important because it is unlikely that organisms, such as pelagic fish within large schools, have the cognitive or sensory capabilities (or information available) to determine their location within their group (which may consist of hundreds of thousands of individuals) and then adjust their position relative to that. I then consider the distribution of fish over larger spatial and temporal scales, and show how deliberately simple self-organising models can help us understand the composition of wild fish schools, and how this relates to the fission-fusion dynamics of populations. I conclude by combining an empirical and theoretical approach to consider how the theory of optimal group size choice in fish can be considered from a self-organisation perspective.

Associative behaviour is defined as the spatial relationship between an animal (or a group) of a given species and another animal of another species, which is based on a decision by at least one of the two individuals to maintain contact with the other associate, but not for the purpose of feeding on the other. This definition is expanded to include objects and topographic structures (seamount, island, etc) that are not the exclusive habitat of the associated animal. Associations between fish and objects, mammals, topographic structures and other species of fish are reviewed, and the different hypotheses for these associations analysed. Finally, a generalisation of the meeting point hypothesis, fitting all types of association, is proposed. This hypothesis, initially applied to the association between tuna and floating objects, proposes that fish make use of animate or inanimate targets to increase the encounter rate between isolated individuals or small schools and other schools in order to constitute bigger schools that are more efficient to the survival of the species. It is strongly recommended to move from the scientific description of the association to specific experimental studies aimed at identifying ethological processes. This would provide a better understanding of the dynamics of associative behaviour of pelagic species, which in turn would permit better stock assessment and fishery management.

Tropical tunas and other pelagic species are often found in association with floating objects. The underlying processes of these associations are still not known. For instance, we still do not know how far fish can detect FADs, what sensory systems they use to detect them, if they use any kind of spatial memory, whether they can orientate towards them or not, how long they stay around FADs. I will review these different questions and show some preliminary results, and discuss protocols for the future.

Small pelagics, such as Clupeids, Engraulids, Carangids, are usually "obligatory gregarious", i.e., they spend all their life inside collective spatial structures, such as schools. These structures present contradictory and paradoxical characteristics: on the one hand they have no evident permanent morphology, and are usually amoeboid and variable in shape; on the other hand, their morphometric features have been proved to be species dependent and allow a statistical identification. This means that there is always some organization in a school. Another indication of this organization has been given by the results of variogram calculations on the density distribution of fish inside the schools as recorded using 3D acoustic data. The relations between the length and width of a school show a permanent allometry, with different ratios according to the external conditions, as observed comparatively in different areas for the same species.
The internal structure of fish schools have been studied using visual observations and multibeam sonar recording. Two types of structures were recognized: vacuoles, that are motionless volumes inside the school where no fish enter and which remain empty while the school is surrounding and passing over them; and nuclei, which are dense volumes encountered in almost all the schools, with often fixed dimensions (around 10 m diameter) independent from the school overall volume.
All these structures are likely resultant of elementary behavioral reactions of the individuals. The allometry between the horizontal dimensions of the school are depending on the school motion, the distance to the source of stress when it exists and the "conflict" between a reaction of the individuals to swim away from the source of stress, and a tendency to maintain a parallel route with the other individuals. Vacuoles are created by discontinuities in the schools, producing small empty areas which represent "frightening volumes" that the fish is avoiding while swimming in the global direction of the school. Nuclei are the compromise between a general aggregative pattern of the individuals, which tend to maintain a fixed position relative to the others (e.g. some body length distance), and a destruction of this pattern at large distances, above which fish cannot maintain synchronous movements. The permanent reshaping of the schools around these two types of structures explains the apparently contradictory results of a 3D variability and the statistically permanent morphology of the school.

Schooling is a common feature of many marine pelagic fish, and has attracted much research on its function, mechanisms and effects. Unfortunately, there is still much uncertainty about social interactions within schools, in particular regarding school fidelity and genetic relationships among school members. Microgeographic genetic differentiation observed in many genetic studies on marine pelagic fish suggests non-random distribution of genotypes among sampled assemblages, and thus potentially some degree of school fidelity. Recent advances in molecular methodology, in particular the development of hypervariable genetic markers (microsatellites), now allow the estimation of relatedness among individuals and the identification of families within fish schools.
Published data on tuna and other pelagic fish can be used to assess the power of such approaches. Simulations suggest that microsatellites would be able to detect relatedness higher than random even if full siblings constitute only 10% of the school. Furthermore, it would be possible to identify pairs of full siblings from microsatellite data. The power of molecular approaches could be further improved by targeted sampling (specific non-associated schools, different life-history stages) and combination with phenotypic markers providing insights in individual environmental history (morphometrics, otolith microchemistry).

Associative behavior is a major component of the life of pelagic fish. Tropical tuna and other species are often found in association with floating objects, with other animals, and with topographic features such as seamounts. Because fish are more abundant, occur in bigger schools, or are easier to catch when aggregated, fishermen extensively use these associations to increase their catches. Currently, around half of the worldwide tropical tuna catch is made from schools associated with floating objects and the trend is increasing. There is a clear need for new technologies to study these associations, and to propose new innovative methods of management. Two agencies (the European Commission and the Pelagic Fisheries Research Programme) have just decided to fund 3-year projects to specify the specifications of future instruments and to develop prototypes. The objective is to define specifications and develop prototypes of (1) autonomous buoys equipped with 360º sonars to observe aggregations, data loggers to detect individuals carrying electronic tags and satellite uplinks for both, (2) new electronic tags with ecological sensors. Various field experiments will be conducted to assist in development of specifications for the autonomous sonars (acoustic tags, acoustic surveys, …). The new instrumented buoys will become observatories of pelagic ecosystems. They will reduce the dependence on research vessels and represent a major advance in fisheries research and the study of pelagic organisms.

Biological interactions occur at specific locations involving the spatial redistribution of organisms. From initially homogeneous states, striking heterogeneous spatial patterns can emerge. Recognizing the importance of space, biologists have struggled with the difficulties of collecting data across spatial scales.
Recent advances in ocean monitoring technology are generating new insights on how fish move and interact. Mathematical modeling facilitates the exploration of the complexities in observed dynamics and aid in addressing hypotheses difficult to test in laboratory studies or open-ocean experiments.
A spatially-explicit, individual-based model of bluefin tuna whereby interacting individuals coexist on a spatially heterogeneous ocean landscape will be presented. The model represents the population dynamics of schooling bluefin tuna seasonally resident in an important Northwestern Atlantic commercial fishing area. The model is structured based on new results obtained from the analysis of acoustic tracking, satellite tagging and survey data. With continued refinement and improvement, the model may lead to predictions of emergent patterns of bluefin spatial distribution, whereby spatial patterns are characterized on the basis of individual decision-making in schools, seeking to maintain survival and improve evolutionary fitness. Selected results from analyses of new experimental data and model simulations will be presented. The talk will end by outlining several improvements required and two aspects of the model where further research is continuing.

During the last decade the technological developments and improved spatial and temporal resolution in hydro-acoustics, have made it possible to study schooling and aggregation behaviour of fish and predator-prey interactions in considerable detail. However, a fundamental challenge is still to be able to track individual fish within a school or aggregation of fish at the same time as we observe the collective movements. We need to find the underlying mechanisms and motivations of individual fish swimming in aggregations, in order to improve our understanding and predictability of collective behaviour. Focus should be put on individual fish behaviour within schools and aggregations. We need proper scientific tools and methods to reveal how individuals behave within aggregations under natural and variable conditions in the field. In future studies, we should synoptically combine echosounder and multi-beam sonar studies with individual tagging (satellite and acoustic tags) and detailed underwater video-observations, which also could provide useful inputs to individual-based fish school models. The spatial resolution of hydro-acoustics is usually not high enough to track individual fish within dense aggregations, and we need better post-processing programs for quantitative analysis of multi-beam sonar data. High-resolution underwater cameras and Crittercam-technology will offer new and complementary data sources to our understanding of detailed schooling and aggregation behaviour. We need to test specific clear-cut hypothesis and design field studies accordingly. I will present experiences and provide acoustic and visual examples on behavioural and ecological studies on schooling herring in the Northeast Atlantic. Suggestions of new directions and extensions of fieldwork on schooling and aggregation behaviour of pelagic fish will be discussed.

Fish aggregating devices (FADs) have been enormously contributed for tuna fisheries in tropical and subtropical regions. However we still do not know why/how tuna aggregate around FADs. The main objective of this study is primarily to understand how tuna behave around FADs rather than why tuna aggregate around it by combining various methods of long and short term monitoring of the tuna behavior. So far I tried following researches; 1) long term monitoring with automated acoustic receivers and coded ultrasonic transmitters, 2) ultrasonic tracking, 3) tagging with the archival tags. I will show primarily results of automated monitoring experiments with acoustic devices and discuss the behavior around FADs. We released 54 yellowfin and 8 bigeye tuna tagged with coded ultrasonic transmitters around 7 monitoring FADs, which were deployed around Okinawa Islands. We could monitor 95 % of tagged tuna. Most of them stayed near the FADs continuously without a few hours absence for certain period. The continuous residence time (CRT) was defined as duration that a tagged tuna stayed near the FAD continuously without a day scale absence. The maximum of CRT for yellowfin and bigeye tuna were 55 days and 34 days, respectively. Furthermore 50 % points of cumulative survival rate of these CRTs were estimated 9.9 days and 7.0 days, respectively. There was no significant difference between yellowfin and bigeye tuna. Although most of tuna stayed around FADs continuously, frequency of signal detections per hour showed approximately 24 hours periodicity for 70 % of yellowfin tuna and all of bigeye tuna. These periodic fluctuations of detections may reflect changes of horizontal distribution of tuna around FADs rather than of vertical distribution, and were categorized 5 patterns. Furthermore these periodic patterns may imply that tuna behave with regularity around FAD as the base point of aggregating behavior and they could keep recognizing the location of FAD.

Early work on schooling behavior focused on the question "How" - how do fish manage to react and interact with such a high degree of coordination? Studies sought to project individual fish onto rigid lattice structures, align them using optimotor responses, or imbue them with an egalitarian sensibility. With the rise of evolutionary biology and ecology, many of these approaches fell out of favor, in part because group response was often synonymized with group selection. Instead, studies began to focus on the question "Why" - with the answer implicitly at the level of the individual. The vast majority of behavioral papers on fish schooling address either predation, or competition for food resources, as the agents of natural selection structuring schooling behavior. In the last decade, four major events have simultaneously moved schooling research both backwards and forewards. First, computational abilities have vastly increased, allowing simulations and dynamic models of schooling not possible in earlier decades. Second, humans - the ultimate predator - have targeted schooling fish in world fisheries. This has lead to a body of work on how behavioral approaches can be used to fish - and to protect - schooling species. Third, the development of complexity theory and the rise of agent-based models have provided for a re-examination of the How question. Finally, group selection is now being reconsidered as part of multi-level selection, in which both the individual and the group can play active roles in species evolution. What does all of this mean for schooling fish? Where should research be directed in the future? Fish schools are an excellent biocomplexity model: Individuals, using only local knowledge (I sense you, I react to your movement), interact to form group responses (now known as epiphenomena or emergent properties). At the same time, our evolutionary and behavioral ecology models predict that these responses must confer advantage - or at least not disadvantage - on the individuals performing them. Thus, fish schools continue to be an excellent proximate evolutionary model. Finally, because many of the schooling species in the marine environment are so prone to overfishing because of this very behavior, fish schools are an excellent model of the degree to which measured human disturbance can upset a system based on both complexity and selection. Thus, the future of fish schooling behavior lies - in my opinion - with marrying the questions How and Why, as well as the approaches of complexity with multi-level selection, and doing all of this before human pressures have so massively altered schooling fish populations as to render all of our research moot.

One of the most noticeable features of individuals in a single shoal is the relative uniformity in size, shape and colour. Moreover fish shoals tend to be formed by individuals having the same physiological state such as hunger. It is possible that the size and shape of the shoals in the wild reflects fish’s decision to stay or leave a group of fish. How fisheries impact on fish shoaling behviour, particularly in the homogeneity of fish shoals is not well understood. In this meeting, I will present some theoretical assumptions on how purse seine fisheries may affect the structures of small pelagic fish shoals which may contribute to the collapse of this fisheries.

A school based model is briefly presented. Detailed annual spatial distributions of fish density were derived using geostatistics from 2.5 n.mi. integrator data from 13 years of acoustic surveys of herring in the Orkney Shetland area from 1989 to 2001. The annual spatial were used to derive a generic underlying spatial probability distribution for North Sea herring for the month of July. Data from a subset of surveys were used to define distributions of school size and school MVBS. Changes in school size and distribution with time of day were taken into account using relationships derived from an image analysis database of school characteristics. The spatial distributions of different size classes of schools were modelled separately where differences in distribution were found. Interactions between school size and density were included. The systematic motion of the population was described by horizontal migration based on 40 years of tagging data. In addition, local motion was added based on sonar observations and on school swimming speed. The motion and spatial distribution were modelled using a method due to Metropolis, modified to include migration in addition to the spatial preferences and random motion. The model provides a description of the population at an individual school level for a short period of time, (1-2 months). The current data needs, the short comings and the information required to provide a more realistic model are discussed.

Although it is accepted that the behavior of a fish school is determined by internal (e.g. physiology) and external (e.g. environment) forces, there is a clear lack of knowledge on the role of individual characteristics on schooling behavior. Through the study of some behavioral parameters (inter-individual distance, distance to nearest neighbor, trajectory, space occupation, ...) on captive fish, we examine the influence of individual characteristics of pelagic fish (species, sizes) on the cohesion of schools. We used outside circular tanks to keep alive individuals during several months. The observations consisted in 5-mn video sequences that were analyzed in the laboratory via post-video processing. Three species (Kuhlia mugil, Selar crumenophtalmus and Decapterus macarellus) have been studied, with 6 different fish densities. The first species (Kuhlia mugil) is considered to be an "optional schooler", living in groups on the outer slope of the coral reefs, while the two last species are considered "obligate schoolers". Each species was studied with mono-specific schools, and the two last species ("obligate schoolers") were also studied mixed together. Experiments with different sized fish were also conducted. The preliminary analyses on the first species (Kuhlia mugil) shows interesting behavioral variations as responses of individuals to various fish densities, giving new insights to the optional gregarious character of this species. The comparison with the two other species ("obligate schoolers") tend to show that each species might have a "gregarious index" (optional vs. obligate), which "fixes" the capacity or the incapacity of individuals to form a coherent and stable group. Therefore, following these first results, individual sizes and/or species seem to be less crucial for the formation and cohesion of fish schools than the "gregarious index" of the individuals forming the school.

Aggregate and group behaviour in fish is often seen as an example of a complex collective behavioural phenomenon. The most dominant collective behaviour of fish is schooling. Since more than 4000 species of pelagic fish are schooling, the mechanisms and dynamics of this behaviour is important to understand. One way to understand a natural phenomenon is through reconstruction and simulations, and to compare the behaviour in the model with the corresponding natural system. I will present an individual based model, operating in a continuous 3-D space, where individual fish is represented as independent agents, with local perception restricted by distance and angle of view. Based on perception information, specific rules are used to calculate the behavioural response for each fish. Rules representing interaction between neighbouring fish includes repulsion, attraction, velocity and acceleration match, while additional rules are used for other interactions such as predator, food and the environment. In the simple rule model system, the rules are defined precisely and given an identical configuration of the surrounding, the same response to the fish will be produced. In a more dynamic model, rules are tuned according to a "situation state parameter" (inside school, outside school etc.). I will also show how neural networks can be trained to replace static rules. The models can be used to demonstrate how different combinations of rules will result in different collective behaviour similar as observed in natural fish schools including milling, tight ball, herd, split and join. However, similarities between the natural system and the model system are mainly reflected in the collective patterns, and the biological relevance of perception, individual rules and interaction parameters are still incomplete. An important experience obtained from trying to build realistic models of schooling fish, is thus the need for more precise knowledge from field, laboratory and behavioural studies.