Two resolution pathways emerge from the sequence analysis. The stronger is lethal, running from gunfire to the killing of a victim. The second is economic, running from a ransom demand to payment and release. Ransom is a coherent pathway but not the organising logic of the phenomenon.
Kidnapping in Nigeria is often described by its causes and its costs, but rarely by its internal order. This report reconstructs the event itself: the chain of behaviours through which an abduction moves from first contact to release, death or rescue. Behaviours were coded from the full text of media-reported incidents, ordered by a canonical event model, and analysed with lag-one behaviour sequence analysis. The result is a behavioural map covering more than fourteen thousand reported incidents. The behaviour set was derived inductively from what reports describe rather than from an assumed ransom script, because the underlying corpus was collected on kidnapping, abduction and hijacking terms, with no ransom keyword. Two resolution pathways emerge. The stronger is lethal, running from gunfire to killing. The second is economic, running from a ransom demand to payment and release. Ransom is therefore a real pathway, but one among several rather than the organising logic of the phenomenon.
Existing scholarship on kidnapping concentrates on what drives it, what forms it takes, and what it costs. What that literature does not do is reconstruct the within-incident order of behaviours through which a kidnapping actually unfolds.
Accounts of causes and consequences explain why kidnapping occurs and what it produces. They do not establish how an incident progresses, or which behaviours reliably follow which. Behaviour sequence analysis addresses that gap. Rooted in Markov models, it treats an offence as a chain of discrete acts and estimates, for each pair of behaviours, how likely the second (the sequitur) is to follow the first (the antecedent), then tests that likelihood against chance. The output is not a story of motive but a map of structure.
The map is built from the full text of media-reported incidents drawn from Nigerian newspaper and broadcast archives. Each report was scanned for a fixed codebook of behaviours, from the arrival of armed men to the discovery of a body, and the behaviours present in each incident were assembled into an ordered sequence. A transition frequency matrix records how often each behaviour is followed immediately by each other; a global chi-square test confirms that the observed transitions depart sharply from independence; and per-cell standardised residuals identify which specific transitions carry that departure.
The standardised residual is the unit of interest throughout. It measures how far an observed transition sits above what chance alone would produce: the larger the residual, the more reliably that behaviour follows its antecedent. In the diagrams below, arrow thickness is banded to the standardised residual, so the heaviest arrows are the most dependable steps in the sequence.
The opening of an incident is dominated by the behaviour of armed perpetrators, and its structure is strikingly consistent. Assailants arrive, invade or waylay, discharge firearms to assert control, and move quickly to remove the victim. Where the abduction is from the road, the waylaying of a vehicle is followed by seizure at very high reliability; where it is a community raid, arrival is followed by gunfire and then abduction.
Two features stand out. Firearms are the organising instrument of the opening: arrival, gunfire and abduction form a tight, high-residual chain. And the sequence is opportunistic in its settings but orderly in its conduct: whether the victim is taken from a road, a home or a vehicle, the same compressed routine of subjugation and removal recurs. Contact with the victim's family sits at the boundary of this phase and the next, and is the point from which the demand pathway, where it occurs at all, proceeds.
Once a victim is held, the incident resolves along one of two dominant pathways, and the diagram makes their separateness visible.
The economic pathway is coherent but narrower than the literature assumes. A ransom is demanded, then paid, after which the victim is released (SR = 56.9 and 80.3). That this chain emerges at all is meaningful, because nothing in the coding imposed it. But a ransom demand is recorded in only 17.5% of reports, so the chain describes a substantial minority of incidents rather than the standard case. Rescue by security forces, escape, and death are parallel resolutions of comparable or greater weight.
The lethal pathway is the stronger of the two. Gunfire is followed by the killing of a victim at the highest residual in the entire matrix (SR = 84.1), and the discovery of a body frequently marks the first point at which the state enters the sequence. This asymmetry is the central analytical point of the resolution phase: the state most often responds to an outcome it did not shape. Police behaviour forms a distinct, largely self-contained track, in which notification is followed by pursuit and, in a share of cases, arrest.
The table lists the transitions with the highest standardised residuals, the reliable joints of the sequence. Each is read as antecedent then sequitur, with n the observed frequency and SR the standardised residual.
| Antecedent | Sequitur | n | SR |
|---|
At this scale the behavioural map is not a single archetype but a distribution. Killing is the single most frequently reported behaviour, present in 43.4% of reports, ahead of release at 40.5%. An explicit ransom demand appears in 17.5%. Reading the corpus without a ransom frame also surfaces behaviours a kidnap-for-ransom script omits entirely, notably sexual violence, present in 6.0% of reports, and forced marriage in 2.1%. The geography is concentrated rather than uniform, weighted toward the north, and the settings of abduction are dominated by roads, communities and, distinctively for Nigeria, schools, whose mass abductions account for a large share of the total number of victims even though single-victim incidents are more numerous.
Three limits bound these findings. The sequences are reconstructed from media reports, so cases that are never reported, including those resolved by quiet ransom payment, fall outside them; the map describes reported kidnapping rather than all kidnapping. Behaviours are detected by textual pattern and ordered by a canonical model rather than coded by hand for each incident, which trades some per-case precision for coverage at scale. And the analysis is cross-sectional and lag-one: it captures which behaviour follows which, but not duration, nor how these patterns are shifting over time. The natural next step is a longitudinal, time-structured sequence analysis that can track how the tactical repertoire evolves as security policy changes.