Cheryl Walker, Hostages in Republican Rome
List of Abbreviations
1: Meaning and Purpose of Hostageship in the Greco-Roman World
2: The System of Hostage Regulations in Rome and the Greco-Roman World
3: Roman Conduct toward Foreign Hostages
4: The Termination of Hostageship
Appendix II A
- Cheryl Walker, Hotages in Republican Rome, Chapter 2: The System of Hostage Regulations in Rome and the Greco-Roman World [28-79], http://chs.harvard.edu/publications.sec/online_print_books.ssp. Center for Hellenic Studies, Washington, DC. 2005
Chapter 2: The System of Hostage Regulations in Rome and the Greco-Roman World [28-79]
The formal regulations concerning hostages reveal considerable variety in detail. Nonetheless, it is worthwhile to examine the several categories into which the regulations may be classified in order to determine the underlying principles by which hostages were chosen and to attempt the reconstruction of a typical case. I have used material from Greek and other Mediterranean cultures as touchstones, so to speak, for the Roman incidents, so that both traits in common and features which are peculiarly Roman may be isolated and appreciated;  although the difficulties in such an approach are manifest, it will help to prevent in some measure the distortion which would be inevitable if one considered Roman hostage exaction without taking into account the broader context of the Greco-Roman world. The major problem, of course, is the paucity of examples to be found in a given category, for, of the many occasions on which hostages were exacted, there are relatively few for which any further details are attested (63.3%). Often, too, it is the extraordinary nature of the details which has prompted their documentation, giving the reader a distorted view of the entire system.  We will therefore need to bear in mind these caveats, as we examine the conditions stipulated by treaty or other formal agreement.
One question, however, must be posed and tentatively answered before we survey the available data on hostage requirements. Whose  practices and preferences had precedence? In a legal contract, both parties agree to the terms, but a nation desperate for peace may have no meaningful choice. It has been asserted that the recipient state determined the specifics of sex, age, number, and affiliation, and some ancient testimony supports the idea (e.g., Polybius 29.3.4; Caesar Gallic War 5.4.1-2)  . While this hypothesis is probably not always true, it is likely to be correct for the majority of cases. On the other hand, the recipient of hostages would frequently have had customs not too dissimilar from those of the donor, as the two were often neighbors and from the same racial and linguistic stock, or, when practices differed, the recipient might well have respected native custom in order to minimize resentment. 
Two facts have obscured the question of the preferred gender for a hostage: first, the linguistic bias which permitted a group of both men and women to be collectively described as male, and second, the unspecified gender of the words for hostages and children. For the purposes of this study, I have assumed a neutral value for such words as πα?ς, παιδ?ον, τ?κνον, and liberi, and a masculine value only where the male gender seems indisputable (e.g., ?νδρες and similar words, or where military service is indicated). This assumption is perhaps overly conservative and severely limits the examples under discussion, but it seems to be the only adequate guarantee of an avoidance of circular reasoning. Even a cursory glance at Table I discloses a marked preference for male hostages over female or mixed groups. With the exception of  legendary material,  the earliest examples of female hostages date from the
TABLE I: HOSTAGE GENDER IN HISTORICAL EXACTIONS
|Type of Exaction|
|Nationality of Recipient State|
|Male||1 (2.3%)||17 (38.6%)||10 (22.7%)||8 (18.2%)||0 (0.0%)||36 (81.8%)|
|Female/mixed||0 (0.0%)||3 (6.8%)||4 (9.1%)||0 (0.0%)||1 (2.3%)||8 (18.2%)|
|Male||0 (0.0%)||13 (41.9%)||8 (25.8%)||7 (22.6%)||0 (0.0%)||28 (90.3%)|
|Female/mixed||0 (0.0%)||2 (6.5%)||0 (0.0%)||1 (3.2%)||0 (0.0%)||3 (9.7%)|
|Male||1 (11.1%)||3 (33.3%)||0 (0.0%)||0 (0.0%)||1 (11.1%)||5 (55.5%)|
|Female/mixed||0 (0.0%)||0 (0.0%)||0 (0.0%)||3 (33.3%)||1 (11.1%)||4 (44.4%)|
|Total||2 (2.4%)||38 (45.2%)||22 (26.2%)||19 (22.6%)||3 (3.7%)||84|
 fourth century B.C., although the date of the exact incident which may claim the dubious distinction of the first such exaction is a vexed problem. According to Plutarch, Antipater had demanded fifty children from Sparta in 331 B.C., after his defeat of Agis; the ephor Eteocles, fearing that the children would be uneducated because they were not trained in the traditional manner, offered instead twice the number of old men or women (Plutarch Moralia 235 B-C). Two factors make this anecdote suspect. Had the Spartan counterproposal occurred, the combination of counterproposal (which is not found in the other sources for this episode) and of the first Greek exaction of female hostages ought to have merited comment before Plutarch’s time. Moreover, the four centuries which had elapsed between Antipater’s demand and the time of Plutarch had witnessed a significant change in Greek gender preference for hostages,  and Plutarch could easily have inserted this anachronistic detail in an anecdote or simply have accepted it from his source, who had made this kind of error. In any case, the late and unsupported testimony of Plutarch is insufficient to permit the conclusion that female hostages were a feature of Greek political negotiations as early as 331 B.C.
In 326 B.C., Alexander of Epirus sent three hundred familiae from southern Italy to Epirus (Livy 8.24.5). The statement that entire familiae were exacted is very curious and unparalleled: it is not obvious what individuals are understood by the term, but one can readily assume the presence of women among them. As Livy, the authority for this story, is also a late source, and one especially  prone to contamination from unreliable sources, it is difficult to accept the full historicity of this incident, and it is more likely that it is the earliest known exaction of females by Greeks.
Nevertheless, the custom had certainly appeared by 303 B.C., the year of the Metapontine campaign of the Spartan Cleonymus and the Lucanians. Duris of Samos is cited as the source for the parenthetical observation that Cleonymus was πρ?τος ?νθρ?πων ε?ς ?μηρε?αν λαβ?ν παρ? Μεταποντ?νων γυνα?κας κα? παρθ?νους.... (Athenaeus 13.605D-E). This statement cannot be true in its absolute sense. Even if we discount the demand of Antipater related by Plutarch and the exaction attributed to Alexander of Epirus, we cannot accept the information as it stands, for the Egyptians and the Persians had taken female hostages at least two generations before.  The simplest solution, if we attempt to rehabilitate Duris’ remark, is that something like ?λληνικ?ν either has fallen out of the text or was what Duris meant but did not write.  Whether our difficulty with the phrase πρ?τος ?νθρ?πων is resolved by textual emendation, by a narrowed understanding to be derived from context, or by merely declaring Duris in error on this point, it is clear that the Greeks exacted female hostages by formal international agreement at least from the time of the late fourth century B.C.
What is astonishing is the speed with which this custom gained acceptance in the Greek world. In 361 B.C., Rheomithras had given his wife, his children and the children of his friends as hostages to Tachos, the king of Egypt (Xenophon Education of Cyrus 8.8.4; Diodorus 15.92.1); Xenophon coupled Rheomithras with Mithradates, who had betrayed his own father, and used them both as examples of evildoers who had profited by their  immorality. Obviously, such conduct was despicable to the conservative Xenophon, whose works probably reflected the sentiments of many Greeks of the early fourth century. The disapproval with which exaction of female hostages was viewed had not disappeared by the end of the century, since the motives attributed to Cleonymus in 303 B.C. are those of lust and sexual indulgence (Diodorus 20.104.3).  Yet an anecdote attributed to Agesipolis the son of Cleombrotus (probably Agesipolis III of Sparta)  indicates that by the second half of the third century B.C. women and children were the usual hostages: when someone said that Agesipolis while king had served as a hostage with those in the prime of life, and not their wives or children (ο?χ ο? πα?δες ο?δ? α? γυνα?κες α?τ?ν), Agesipolis declared it fitting that they bear the consequences of their own mistakes (Plutarch Moralia 215 BC.). Although Plutarch’s authority for the historicity of this event is justly suspect, other incidents from this period support the tone of Plutarch’s anecdote. In 223/2 B.C., Ptolemy III Euergetes’ demand for Cratesicleia, mother of the Spartan king Cleomenes, was not an outrage, but merely a cause of shame to Cleomenes (Plutarch Cleomenes 22.3-7).  So much had opinion altered that Polybius, in the second century B.C., could criticize the Carthaginians of 241 B.C. for not keeping the mercenaries’ wives and children as extralegal hostages (Polybius 1.68.3); in another passage, which describes in a gnomic manner the best pledges for the future integrity of a person, Polybius listed wives, children, oaths, and previous behavior (Polybius 8.36.3).  Female hostages had apparently become a standard feature of Greek practice. 
Nor was female exaction confined only to Greek spheres of influence. As mentioned above, Rheomithras offered his relatives and those of his friends as surety for his loyalty to the king of Egypt in 362/1 B.C.;  in 334/3 B.C. the Greek general Memnon delivered his wife and children to the Persian king Darius, and, like Rheomithras, was rewarded with the supreme command of the king’s forces (Diodorus 17.23.5). Both of these instances were private agreements and involved non-Greek monarchs, but it may be that, among the Persians, female hostages guaranteed formal treaties as well; on one occasion, Egyptian wives and children remained in Persian custody while the Egyptians acted as guides for Greek mercenaries in Persian pay in 350/49 B.C. (Diodorus 16.48.3).  Alexander the Great is alleged by a late and unreliable source to have received hostages of both sexes on his Indian campaigns in 326 B.C. (Polyaenus 4.3.30). The Carthaginians also took female hostages, for, when Scipio captured Cartagena in 209 B.C., he discovered both males and females of a variety of ages among the Iberians detained there (Polybius 10.18.6; Livy 26.49-51.1; 27.17.1-3;16; etc.). 
Roman practice contrasted sharply. Only four episodes concerning female hostages are attested after 389 B.C., when Roman accounts may be said to begin to be substantially historical. Of the four episodes, two occur in the context of factional infighting: the planned seizure of Pompey’s children (one a girl) in 63 B.C. by the Catilinarian conspirators (Plutarch Cicero 18.1) and the attempted detention of Octavian’s mother and sister by the Republicans in 43 B.C. (Appian Civil Wars 3.91). In neither case was the sex of the hostage a meaningful  factor in the selection; the relationship with the leader of the other faction alone seems to have determined the choice of hostages. Thus there are only two passages which suggest actual Roman demands for female hostages: the possible existence of women hostages taken during Pompey’s Eastern campaigns (Appian Mithridatic Wars 103; 117),  and Augustus’ policy toward some barbarian tribes (Suetonius Augustus 21. 2).
Only the latter may be regarded as indisputably historical, and it is of considerable value for understanding Roman custom, its efficacy, the flexibility of Augustus’ foreign policy, and foreign customs. Let us examine the pertinent passage in full.
. . .tantumque afuit a cupiditate quoquo modo imperium vel bellicam gloriam augendi, ut quorundam barbarorum principes in aede Martis Ultoris iurare coegerit mansuros se in fide ac pace quam peterent, a quibusdam vero novum genus obsidum, feminas, exigere temptaverit, quod neglegere marum pignera sentiebat; et tamen potestatem semper omnibus fecit, quotiens vellent obsides recipiendi.
[. . .and so far was he from a desire for increasing in any way the empire or his military reputation, that he compelled the leaders of some barbarians  to swear in the Temple of Mars Ultor that they would abide by the agreement and peace which they sought, (and) from some barbarians he tried to exact a new type of hostages, women, because he felt that they disregarded males as pledges; nevertheless, he always granted to all permission to replace hostages as often as they wished.]
Four points must be emphasized: 1) the exaction and the oath-taking are evidence of Augustus’ peaceful objectives;  2) women were a “new type” of hostage; 3) his innovation was prompted by the failure (in his opinion) of the traditional male hostages;  4) Augustus mitigated the apparent severity of his innovation by allowing frequent exchanges of personnel,  again a proof of his lenient disposition.
It is the second point with which we must here concern ourselves. According to Suetonius, Augustus did something unusual in  choosing female hostages, and, since there has survived only one other possible historic case of women being detained as formal hostages at Rome, we should accept Suetonius’ testimony as substantially accurate. Thus the available evidence indicates that during the Republic the Romans habitually selected male hostages, and that female hostages were rare exceptions.  Despite the prevailing Greek practice in the second century B.C., when contact with the, Roman political theory no alteration in this criterion for hostage selection occurred until the Romans met peoples who did not seem sufficiently restrained by the threat of death against their male hostages. Only then was the ineffectual Roman practice discontinued in favor of a custom perhaps native to those on whom Rome inflicted it.
The restrictions on age and the references to the age of particular hostages are perhaps the details most often related in our sources; superficially, the evidence seems straightforward, but on closer examination one becomes aware of difficulties. References to age may be classified into three groups: 1) an actual age or range of ages is mentioned in the pertinent passage (e.g., the Aetolian Treaty of 189 B.C. (Polybius 21.32); 2) the age of an individual is known or can be inferred from other sources (e.g., the age of Philip, son of Amyntas in 369 B.C. [Diodorus 15.67.4; Plutarch Pelopidas 26.4, plus the knowledge that he was born ca. 383 B.C.]); 3) the age is implied by such descriptive words as πα?ς, τ?κνον, υ??ς, filius, and liberi. It is, of course, the last group which is the  most difficult to comprehend, for
TABLE II RELATIVE AGES OF HISTORICAL HOSTAGES
|Type of Exaction|
|adult||0 (0.0%)||8(19.0%)||5 (11.9%)||8 (19.0%)||0(0.0%)||21 (50.0%)|
|children||1 (2.4%)||7(16.7%)||7 (16.7%)||0 ( 0.0%)||0(0.0%)||15 (35. 7%)|
|mixed||0 (0.0%)||4*(9.5%)||2 ( 4.8%)||0 (0.0%)||0(0.0%)||6 (14.3%)|
|adult||0 (0.0%)||4 (8.7%)||3 (6.5%)||10 (21.7%)||1(2.2%)||18 (39.1%)|
|child||0 (0.0%)||17(37.0%)||3 (6.5%)||3 ( 6.5%)||1(2.2%)||24 (52.2%)|
|mixed||0 (0.0%)||1 ( 2.2%)||0 ( 0.0%)||3 ( 6.5%)||0(0.0%)||4 ( 8.7%)|
|adult||0 (0.0%)||2(15.4%)||0 ( 0.0%)||1 ( 7.7%)||0(0.0%)||3 (23.1%)|
|child||1 (7.7%)||3(23.1%)||0 ( 0.0%)||0 ( 0.0%)||0 (0.0%)||4 (30.7%)|
|mixed||0 (0.0%)||0 (0.0%)||2(15.4%)||2 (15.4%)||2(15.4%)||6 (46.2%)|
|totals||2 (2.0%)||46(45.5%)||22(21.8%)||27 (26.7%)||4 (4.0%)||101**|
*Including the Samian hostages of 441/0 B.C., following the superior testimony of Thucydides 1.115.3.
**I have omitted two incidents. The Volscian hostages of 495 B.C. are called liberi in Livy 2.22.2, but ?νδρες in Dionysius of Halicarnassus 6.25.2, and the alternative plans of Plutarch Moralia 235 B-C, seemingly resolved in favor of the Spartan counterproposal, are contradicted by Diodorus 17.73.5. 3
 these words signify a relationship, not a true age classification.  The impulse to think of individuals so described as pre-adult must be curbed as misleading, although in some cases this was undoubtedly true.  Nevertheless, one important inference can be drawn from the employment of these terms: the emphasis on the relationship indicates the hostages’ dependence on another’s authority and their lack of status in their own right. The distinction between “children” and “adults” which I have established is, therefore, less a matter of chronological age than of personal authority, and this peculiarity must be borne in mind in what follows. As will be seen, it is a useful and valid distinction. 
From Table II it is apparent that there are distinct differences in Greek and Roman practice. The most significant of these is the marked preference for child hostages among the Romans, 52.2% of the total cases for such age is specified, as compared to only 35.7% among the Greeks. Moreover, the Roman preference is most evident in formal international agreements, where the potential hostages might come from the most diverse age groups, and the recipient’s preferences could therefore have fullest scope. (An army in the field probably could not have surrendered women or children very easily, and the receiving party would have had to accept hostages from available personnel.) While Greeks accepted adult, child, and mixed groups of 25 hostages for treaties without evincing strong preferences,  Romans chose adults in only four incidents (8.7%), one involving women (Augustus’ exaction of barbarian women [Suetonius Augustus 21.2]) and three involving men of military age, evidently as hostages for a truce  rather than a treaty (Livy 43.17.6; Appian Wars in Spain 48; 50):  all somewhat anomalous cases.
The relative preponderance of private hostages among the Greeks probably reflects a custom which the Romans employed in a far more restricted manner; the concept ?μηρε?α referred to a less technical situation in which persons could declare themselves hostages for their statements (Themistocles’ offer to the Spartans in 479/8 B.C., Nepos Themistocles 7.2; Plutarch Themistocles 19.2: Polyaenus 1.30.4; the Pythagorean Archytas and his followers who were Plato’s sureties for the Younger Dionysius’ good intentions, Plutarch Dion 18.2; 20.1; Phocion’s pledge of himself that the Piraeus would remain safely in Athenian control, Nepos Phocion 2,4), and perhaps the instability of internal and foreign alliances (at least compared to the constancy of the Roman Senate) accounts for the greater number of private agreements (Charon’s offer of his son to his fellow conspirators, Plutarch Pelopidas 9.5-7; Moralia 595; the demands for Agathocles’ children by the leader of an opposing Syracusan faction, Diodorus 20.79.4; demands for sureties in foreign alliances, Polyaenus 5.3.4; Plutarch Pyrrhus 31. 2; Polybius 4.9.5; Josephus Jewish Antiquities 13.204-206  ).
In eight instances, the age of an individual hostage or the age range for a group of hostages is known or can be estimated. Philip of Macedon, son of Amyntas, was allegedly delivered to the Illyrians shortly after his birth about 383 B.C. (Justin 7.5.1; Diodorus 12.2.2,4).  The authority for his hostageship at Thebes in the early 360’s is very much better (Plutarch Pelopidas 26.4; Moralia 178C: Diodorus 15.67.4; Justin 6.9.7; 7.5.1-3; Orosius 3.22.2), and we may reasonably assume  that his age was about fourteen to eighteen years during his detention. It is unfortunate that Philip is our only Greek hostage of known age, for it is quite likely that his case is atypical; certainly his relationship to the king of Macedon was a stronger factor in his selection than his age.  Similarly, the surrender of Antyllus, M. Antonius’ son, to Caesar’s assassins at the tender age of one or two years (Livy Summaries 116; Velleius Paterculus 2.58.3; Cicero Philippics 1.1.2; 1.13.31; 2.36.90; Plutarch Brutus 19.2; Antony 14.1; Appian Civil Wars 2.142; 3.15; Dio 44.34.6) and the planned abduction of the equally young Gnaeus and Sextus Pompey and Pompeia in 63 B.C. (Plutarch Cicero 18.1) were due to their connection with powerful relatives, and their ages were probably not operational factors. These were personal pledges, and parents would in all likelihood have served as well as children.
Of far greater significance are the ranges specified in the Roman treaties with Carthage in 201 B.C. (14-30 years old, Polybius 15.18.8 and Livy 30.37.6), with the Aetolian League in 189 B.C. 30 (12-40 years old, Polybius 21.32.10-11 and Livy 38.11.6),  and with Antiochus III of Syria in 188 B.C. (18-40 years old, Polybius 21.42.22  ). The ancient source for all these treaties and for the ages of Demetrius of Macedon and Demetrius of Syria is the reliable and, for the most part, contemporary Polybius. As Aymard has remarked, the accuracy of these details seems difficult to refute, and the small number of hostages exacted and the minimum age limitation imply that the exaction was not intended to weaken the state militarily. 
Both Demetrius of Macedon and Demetrius of Syria were about ten years old at the time of their exaction. Since the surviving  conditions of the treaty with Philip V do not include age restrictions, this fact is easily fitted into the other data at our disposal. But the age of the Syrian prince is another matter; according to the original treaty with Antiochus III, hostages were to be at least eighteen (Polybius 21.42.22 and Livy 38.38.9). This contradiction may be explained in one of two ways: either the original restriction no longer pertained, or Demetrius was a special case excepted from the restriction. Mørkholm argues that the violation of the age restriction indicated Roman instigation for the exchange of Antiochus (later Antiochus IV Epiphanes) for the child Demetrius,  and it is very probable that the royal hostages had special advantages which outweighed their relatively less desirable age; their affiliation was far more important than the disadvantage of their greater youth. 
Thus within the same quarter-century of Roman history the minimum age for hostages ranged from about ten to eighteen, and the maximum age from thirty to forty or forty-five.  What purpose could these variations serve? Several hypotheses could be postulated to explain these slight differences in age: 1) they are random and arbitrary; 2) they represent a Roman attempt to determine the most suitable age for a hostage; 3) they reflect the Roman response to the differing ages at which differing cultures recognized maturity; 4) they resulted from a conscious Roman policy which adjusted a rather general age requirement to suit each individual case.  As should be obvious, one’s preference among the possible explanations is dependent on one’s somewhat subjective analysis of Roman motivation during this period of imperial expansion. Nevertheless, the very existence of the variations allows us to infer that Roman requirements  were flexible and capable of adaptation.
The problems of variation notwithstanding, our major concern should be the reason for the distinct difference in the Roman preference for children  and the Greek preference for adults. Let us consider the merits of each. The exaction of leaders, particularly those of opposing factions, could potentially paralyze a faction’s resistance and might conceivably cause its disintegration into smaller, more easily subdued subgroups in the absence of a strong leader. Should, however, the detention of opposing leaders be protracted,  it is likely that new leaders would arise to fill the power vacuum, and it is not impossible that the new leaders would pose a greater threat because of greater competence. Such new leaders might well perceive the former leaders, now hostages, as personal rivals and feel no distress at endangering them; their value as hostages would thus be lessened to a considerable degree. Moreover, the influence of leaders among their fellow citizens may be partly a function of their maturity and experience. The old men who are especially esteemed for their advice are not good risks for long-term detention, as their greater age and probable poorer health increase the odds against their surviving the period of detention. The selection of children, at least in the sense of those who are not leaders in their own right, obviates the problem posed by the possible emergence of new leaders; the same people remain in control of the state, but are themselves controlled by the fear of harm to their children. However reluctantly, these leaders are compelled to cooperate with the recipient state, or their children will suffer. In this  theoretical context, it becomes evident that the recipient state would require some provisions to safeguard the value of the hostages it exacted, and it is possible to see in the attested minimum and maximum age limits an attempt to secure hostages in the healthiest range of the age spectrum. 
Yet there is another advantage in taking children: the ease with which children assimilate and adopt new ideas and customs. The extent to which the Romans recognized and utilized the receptiveness and malleability of youth has long been debated, and scholars have argued for both complete recognition and complete ignorance of the value of Romanized hostages. Nevertheless, by the early Principate even such derivative intellectuals as Diodorus and Plutarch had observed the impact that a foreign education had upon hostages: we have already seen Eteocles’ refusal to sacrifice the education of Spartan boys (Plutarch Moralia 235B-C); Philip son of Amyntas learned much military strategy from his stay at Thebes during the hegemony of Epaminondas (Diodorus 16.2.2); and Sertorius could entice the Iberians to support him with the promise of sharing his command with the Iberian boys trained in Roman ways (Plutarch Sertorius 14.2).  Here we may note simply that the Romans preferred, under ordinary circumstances, to exact children, in the sense of persons not yet sui iuris, and varying in age from twelve to forty years, and that they took independent adults only when circumstances so demanded.
Despite the notorious difficulties involved in studies based on numbers, due to the ease with which numbers are corrupted in the  textual tradition, a survey of those ancient
TABLE III: NUMBERS OF HOSTAGES EXACTED
|Recipient State||Donor State||Number Exacted||Donor State's Political Structure|
|Athens||Opuntian Locrians||100||Autonomous city-states|
|Macedon||Sparta||50 or 100*||Oligarchic Monarchic city|
|Epirus||South Italy*||300||Autonomous cities familiae|
|Demetrius Poliorcetes||Rhodes||100||Oligarchic city-state|
|Cleonymus & Lucanians||Metapontum||200||Oligarchic city-state|
|Sparta||Achaean League||300||Oligarchic federation|
|“||Cora and Pometia***||300||Autonomous city-states|
|“||Arretium**||120||Subject to Rome|
|“||Carthage||150 or 100*||Oligarchic city-state|
|“||Aetolian League||40||Oligarchic federation|
|“||Cephallenian cities (Same, Cranii, Palensia)||20 each||Autonomous city-state|
|“||Isaura Nova||100||Oligarchic city|
*The figures so marked are given differently in different sources or stated as alternatives for different age groups. For a discussion of these differences, see below, p. 85 n. 48.
**These instances, unlike the others in this table, may not have been exactions specified by a treaty.
***These incidents belong to Rome’s legendary period, and their historicity is dubious.
 passages which attribute definite numbers of hostages to specific incidents may illuminate certain principles of selection. For lack of better evidence, I have assumed the general accuracy of the texts as they stand. The range in the number of hostages as related by ancient sources is remarkable, Octavian exacted 700 Dalmatians in his campaigns of 35-33 B.C. (Appian Illyrian Wars 28), while Antony took only two from Antiochus I of Commagene after the siege of Samosata in 38 B.C. (Dio 49.22.2).  mp;nbsp; Both of these episodes, however, are unusual; for the most part, a range of 20-600 is far more typical, as can be seen from Table III. The factors which established this wide range deserve careful analysis.
One such factor is, clearly, the population of the donor state. Had, for example, 600 hostages been exacted from Sparta in 195 B.C., there would have been severe military repercussions throughout the Peloponnesus, to the advantage of Philopoemen’s Achaean League.  Although the reforms of the last kings and of the tyrants had ameliorated the critical shortage of free Spartan male citizens, the demand for hostages must have recognized Sparta’s depleted manpower. In like manner the relatively few hostages from the Cephallenian cities (Livy 39.28.6) reflect the inability of those cities to provide numerous citizens of the requisite quality. Nevertheless, population was by no means the only or even the major criterion. Such a criterion leaves unanswered the question of why each of the Cephallenian cities should furnish as many hostages as did the Seleucid monarch Antiochus III the Great of Asia or why Sardinia in 176 B.C. delivered 230 hostages (Livy 41.17.3) when Carthage had only one hundred or  one hundred fifty exacted in 201 B.C. (Polybius 15.18.8: Livy 30.37.6; Appian Punic Wars 54). 
The key to the solution lies in the quality of hostages which can be taken, itself a product of the donor state’s form of government. The quality of the hostages is in inverse relation to the number of hostages needed to guarantee the agreement and in direct proportion to the size of the ruling class. The obvious validity of this statement needs no argument. Neither the total number nor the names of the other hostages whom Philip V furnished at the conclusion of the second Macedonian War are known because Demetrius, son of Philip V, was the only one of real importance. The treaty with Antiochus III of Syria regulated the triennial replacement of all the Syrian hostages except Antiochus, the son of Antiochus III (Appian Syrian Wars 39); since Antiochus was the second legitimate son, only the firstborn son and heir apparent Seleucus could hove been a hostage of equal or greater quality. Control of the major families of an oligarchy was correspondingly more difficult, and that of a democracy almost impossible, which may partially explain the Roman preference for oligarchic or monarchic rule in states subject to her. The fact that the type of government was relatively more important than the size of population is evident from the disproportionately large numbers of hostages exacted from the Illyrian Dalmatians (Appian Illyrian Wars 28) and the Sardinians (Livy 41.17.3) compared to the modest numbers demanded from the Aetolian League (Polybius 21.32.10-11; Livy 38.11.6) and Antiochus III of Syria;  the many small, autonomous states of Illyria and Sardinia compelled the exaction of several  hostages from each one, combining to form a far greater total than was necessary for the much larger territory governed centrally by an oligarchy or monarchy.
That the number of hostages is often in round figures - 20, 50, 100 and 300 are those most frequently encountered - should occasion no surprise. The popularity of 300 is more remarkable. As Ogilvie has observed, “300 is the traditional number in legend for hostages.”  Only one instance of 300 hostages in an historical setting appears to have firm contemporary substantiation, and that is of the Carthaginians exacted in 149 B.C. before the outbreak of hostilities (Polybius 36.4.6). In all other cases, the numerical detail occurs in authorities much further removed in time from the events described: 1) the Volscian hostages of 503 and 495 B.C. (Livy 2.16.9, 22.2); 2) the illustrious families sent to Epirus in 326 B.C. (Livy 8.24.5); 3) about 200 B.C., the obscure inhabitants of equally obscure Salmatis (Polyaenus 7.48.1); 4) the Termessan and Numantine hostages of about 143 B.C. (Diodorus 33.16.1), the Numidians of 108 B.C. (Orosius 5.15.7), and the Cretans in 70/69 B.C. (Diodorus 40.1.2-3; Appian Sicilian War 6.1). The last group is particularly suspicious because none of the treaties in which these terms were stated was ratified. It is difficult to imagine why the details of these failed negotiations should survive when the details from legal, ratified documents are so often vague or omitted altogether. It is even more troublesome to explain how a tradition since lost managed to be scrupulously preserved in a late, less reliable source while the extant and more contemporary authority remained silent.  Moreover, the propriety of demanding the same number from the monarch Jugurtha,  the urban and decentralized Crete, and the prosperous but hardly enormous towns of Numantia and Termessus seems rather dubious, especially in light of the criteria of population and type of government.
With the exception of 300, however, there seems little point in emending or discounting the numerical data.  Some of the figures are probably inaccurate, due to either scribal error or authorial imagination.  As neither error nor imagination can be objectively demonstrated, nothing could be gained by denying wholesale the numerical accuracy of our sources, but neither should one assert total acceptance of them. Nevertheless, the inherent logic of the criteria of population and type of government, coupled with the wide range of the received numbers, strongly indicates a flexible and pragmatic policy.
That hostages should belong to the governing class of the donor state is certainly predictable; since the purpose of taking a hostage was to constrain the donor state to obey the wishes of the recipient state, the logical and obvious choice for the hostage was a member of the class which could guide the state into obedience. Naturally, the selection of a monarch’s relative, particularly a son, and of aristocratic family members occurred most frequently, as monarchies and oligarchies were the commonest forms of government. It is significant that individual names are most often associated with royal hostages (e.g., Demetrius of Macedon, Antiochus son of Antiochus III, Demetrius son of Seleucus IV), and, although other hostages were  also exacted at the same time, they were completely overshadowed by the royal relative. The implication is clear: the value of the hostage is proportional to his status. 
The evidence does in fact support the view that hostages normally belonged to the highest rank of the political structure of their state. There are two categories into which this evidence falls: 1) kinship with the kings or factional leaders: and 2) words denoting socio-political standing. The tendency to exact relatives of rulers needs no explanation, since in antiquity kinship bonds promoted both natural affection and religious sentiment.  Unfortunately, the condition of the surviving material does not permit an analysis of the types of relationship preferred, except that children clearly predominate;  in cultures with low rates of life expectancy, parents - who might otherwise have been most commonly the persons who constrained the conduct of the donor -  would frequently have predeceased their children’s mature career. The exaction of brothers (e.g. Philip of Macedon [Plutarch Pelopidas 26.4; Diodorus 15.67.4; Justin 7.5.1; Orosius 3.12.2], the brother of Asander [Diodorus 19.75.1-2], and the brother of the High Priest Hyrcanus [Josephus Jewish Antiquities 13.247]), nephews (e.g., Antigonus’ nephew Ptolemy [Plutarch Eumenes 10.2-3] and Indutiomarus’ nephew [Caesar Gallic War 5.27.2]) and other relatives could have been prompted by the lack or unavailability of closer kin, or by a known hostility to members of the immediate family;  the scanty references to affiliation simply do not tell us the principles by which one or a few relatives were chosen from all the members of the leader’s family.
The second category includes such descriptive adjectives as  πρ?τος, ?πιφαν?στατος, ε?γεν?ς, ?ριστος, ?ξιολογ?τατος, ?νδοξ?τατος, πλουι?τατος, κ?λλιστος, κρ?τιστος, nobilis, illustris, primus, and principes; like the nouns ?ππε?ς and equites, these words are the semi-technical terms for the wealthiest and must powerful citizens of a state, particularly an oligarchy. In the Macedonian monarchy, the body of men known as the “Friends” (?τα?ροι or φ?λοι and amici) sometimes were asked to provide hostages: in 368 B.C., to Thebes (Plutarch Pelopidas 27.3); to Seleucus in 286 B.C. (Plutarch Demetrius 48.1); to Rome in 197-196 B.C. (Polybius 18.39.5-6; Appian Macedonian Affairs 9.2: Zonaras 9.16: Plutarch Moralia 197A: Livy 33.13.14-15), in 172 B.C. (Livy 42.39.6-7), and in 168 B.C. (Polybius 29.4.6). On only two occasions do we find hostages described by a word indicating that they might have been of less exalted status, and those are the surrender of twenty Argive πολ?ται to Cleomenes of Sparta (Plutarch Cleomenes 17.5) and of one hundred Rhodians to Demetrius Poliorcetes (Diodorus 20.99.3); we do not know more.
Which factional leaders were likely to be the targets of an exaction? One might readily predict that a faction which opposed the recipient state would have supplied the greatest number of hostages: it is certainly reasonable to assume that the opponents of the recipient state stood in greatest need of restraint, and that families of proven loyalty to the recipient state were less likely to be compelled to furnish hostages. In theory;  but again the scanty information on the factional policies of individual hostages almost precludes conclusion. It may be noted, however, that the evidence derived from internal, extralegal exaction (Dionysius of Halicarnassus 4.83.3; 6.53.1; 6.84.2;  8.43.4; Livy 24.31.12-13; Diodorus 33.15.1; Plutarch Cicero 18.1; Cicero Letters to Friends 10.17.3; Appian Civil Wars 3.91; 5.137) demonstrates the feasibility of this hypothesis in domestic practice.
With regard to foreign policy, our comprehension of the faction hypothesis of hostage selection must rest upon three instances. In 169 B.C. at an assembly of the Aetolian League, Lyciscus, the leader of the pro-Roman party, urged that the supporters of anti-Roman policies deliver their children to Rome (Polybius 28.4.7); his avowed purpose was the quelling of the recent unrest in Greece. In the next year, Masgaba, speaking for his father Masinissa, requested that Hanno, the son of Hamilcar,  be substituted for another hostage, whose name is lost in a lacuna; he was politely thanked and told that the Roman Senate did not think it fair to exact Carthaginian hostages according to the discretion of Masinissa (Livy 45.145). When the Sequani and the Germans demanded hostages from the Aedui in 61 B.C., all the Aeduan leaders except Diviciacus delivered their children (Caesar Gallic War 1.31.8). Thus factional politics did play some role in determining the individuals chosen as hostages, but probably only if the recipient state had a previously established group of allies within the donor state.
This very tentative observation may have a counterpoise in the exactions made to guarantee armistices. The donor party favoring an armistice seems to have selected the hostages,  and these parties are usually the ones who are already the most closely connected to the recipient state. In a state with two factions nearly equal in strength, such as Carthage in 203 and the Aetolian League in 191 B.C., the armistice party might well have difficulty in forcing the opposing  faction to supply hostages for a settlement which the opposing faction did not find desirable. Moreover, the anti-Barcid party in Carthage and the pro-Phaneas party in the Aetolian League would have been as eager to display their own good intentions as they would have been hard pressed to offer persons from the opposing faction. In these circumstances, it is possible that the hostages came from the party in least need of external control. 
The evidence for who selected hostages is fairly abundant; in 48 cases we are told explicitly which party chose the hostages.  In nine, or 18.7%, of these cases, however, chance played the largest role, for the hostages were seized or detained in an extralegal, usually violent, situation; the rape of the Sabine women (Plutarch Romulus 16.2); the Etruscan prisoners of war held by Tarquinius Priscus (?ν ?μ?ρων ... λ?γ?, Dionysius of Halicarnassus 3.57.1); Mucius Scaevola in 508 B.C. (Dionysus of Halicarnassus 5.31.2); the consular ambassadors at Cumae in 492 B.C. (Dionysius of Halicarnassus 7.2.3; 12.2); the Tarentine envoys detained by Pyrrhus in 281 B.C. (Zonaras 8.2); the residents of Rhegium captured outside the walls by Hannibal in 215 B.C. (velut obsides, Livy 24.1.7-8); the Syracusan army at Leontini in 214 B.C. (Livy 24.31.12-13); the son of the Ilergetes’ chief (Livy 34.12.7); and Philip’s proposed military levy from the Achaean League in 200 B.C. (in loco obsidum, Livy 31.25.8).  A combination of circumstances was responsible for the detention of hostages such as these, and essentially anyone available at the time would have served. Thus it is not meaningful to speak of selection in these cases. 
In ten other cases (20.8%), a recipient state or faction chose extralegal hostages in a more deliberate manner, because of the political position of the hostages themselves or because of a familial relationship to a leader: among the Greeks, Alexander the Great “honored” some Thracian chiefs by taking them with him on his Asian campaigns (Frontinius Stratagems 2.11.3), and Nabis in 195 B.C. arrested his potential rivals (Livy 34.27.7-8; 32.12); among the Romans, very much in the tradition of Alexander, Julius Caesar brought to Britain in 54 B.C. the Gallic leaders whom he believed to be disaffected allies (Caesar Gallic War 5.5.4), and the Pompeians Afranius and Petreius detained the Spanish leaders to prevent Spanish defection to the Caesarian cause in 49 B.C. (Caesar Civil War 1.74.5); following (probably unknowingly) the precedent of Nabis, the Catilinarian conspirators in 63 B.C. (Plutarch Cicero 18.1), the Republicans in 43 B.C. (Appian Civil Wars 3.91), and Sextus Pompey in 35 B.C. (Appian Civil Wars 5.137) attempted to seize targeted members of the opposing faction or their relatives; in 216/5 B.C. picked Campanian nobles garrisoned Sicily, serving both as sureties for Campanian loyalty to Rome and as troops against the threat of Punic and native uprisings, just as Spanish soldiers chosen by Hannibal protected Africa (Livy 21.21.11-13); the outrages which Diêgylis of Thrace inflicted upon the relatives of those who had fled Thrace for the safety of Pergamum were meant as a savage retaliation upon disobedient subjects (Diodorus 33.151). These extralegal seizures and detentions, although they display the recipient’s conscious selection, do not indicate at all the procedure employed in a formal agreement, which is the only sort of exaction in which selection is a negotiable issue. 
In the twelve unilateral exactions involving a treaty, the recipient state chose or approved the hostages in at least eleven.  The sources for the ten Roman instances (including three of dubious historical validity) seem maddeningly vague on the method of selection, as if this were merely a matter of routine reporting;  even where we are told something more, it is sometimes very little (e.g., “from youths between fourteen and thirty years old” of the Treaty of Zama in 201 B.C. [Polybius 15.18.8; Livy 30.37.8], or “freeborn women,” in the Latin demand of 389 B.C. [Plutarch Romulus 29.4; Camillus 33.2; Macrobius Saturnalia 1.11.37; cf. Plutarch Moralia 313A]). In one case, the Roman general was limited, not only by criteria of sex and age, but also by the exemptions accorded to some magistrates ex officio (the στρατηγ?ς, ?ππ?ρχος, and δημοσ?ος γραμματε?ς), and to former hostages to Rome (Polybius 21.32.10; Livy 38.11.6). The former exemption is very like the other Greek incident for which we know who chose the hostages; in 304 B.C., Demetrius Poliorcetes could select one hundred hostages from citizens not in office (Diodorus 20.99.3). The difficulties which might have resulted from the detention of an official by an opposing foreign government are obvious. While no similar conditions are documented for other treaties, it is certainly a predictable limitation and quite plausibly would occur more frequently if complete copies of treaties, and not just summaries, were reproduced by our sources. 
It is difficult to know how a Roman general would have selected hostages in these cases. Did the donor government make suggestions? Two sources indicate that this did happen, and that the Roman commander had the right of approval (the Treaty of Zama [Livy 30.37.6] and the armistice with Nabis in 195 B.G. [Livy 34.35.11]).  Yet the temptation to be rid of worthless persons or political rivals would be enormous to a leader permitted to nominate his own hostage to a foreign power; the quality of such hostage would be minimal.  If one knew enough of the donor state’s political situation, one could circumvent the problem of relying upon information provided by an enemy in whose self-interest it was to lie; one could demand hostages by name. Thus Scipio selected hostages from Antiochus III in 190 B.C. (Polybius 21.17.8), Pompey from Artoces of Iberia in 65 B.C. (Florus 1.40.28; Dio 37.2.6-7), and Caesar from the Treveri in 54 B.C. (Caesar Gallic War 5.4.1-2).
The only bilateral exaction for which we have evidence on the criterion of selection is that of Perseus of Macedon and Genthius of Illyria in 168 B.C. (Polybius 29.3.4,6). Each king received from the other the hostages whom he wished, and it appears that each had a list - and definite ideas about the persons best suited for the purpose.
But the recipient did not always have the ability to select hostages. In private agreements, the recipient sometimes did not possess the physical superiority to enforce his will with regard to his selection criteria and instead was compelled to accept what was offered. When the Greeks Themistocles (Plutarch Themistocles 19.2; Nepos Themistocles 7.2; Polyaenus 1.30.4) and Phocion (Nepos Phocion 2,4) pledged themselves as security for their statements, when Charon volunteered his son (Plutarch Pelopidas 9.5-7; Moralia 595) and Rheomithras (Xenophon Education of Cyrus 8.8.4; cf. Diodorus 15.92.1) and Memnon their entire families (Diodorus 17.23.5) for their loyalty to a cause, and when Agathocles guaranteed a truce  with Ophelas of Cyrene by the delivery of his son (Polyaenus 5.3.4), and when the anonymous Cumaean offered himself to prove the genuineness of his betrayal of his comrades in 492 B.C. (Dionysius of Halicarnassus 7.10.5-6), the recipients could apparently only accept or refuse the proposal; besides, the donors had already offered the most effective types of hostages.  The Romans are attested as recipients in two such incidents, the detention of one of the two men who had declared their willingness to betray Nequinum in 299 B.C. (Livy 10.10.3) and the surrender of two of Masinissa’s retinue for a conference between Scipio and the Numidian prince (Livy 28.35.4).
Two peculiar cases should also be classified as instances in which the donor was able to select the hostages. In 285 B.C. Seleucus I captured Demetrius Poliorcetes; Antigonus Gonatas, Demetrius’ son, offered himself in place of his father (?μηρε?ειν ?τοιμος ?ν α?τ?ς ?π?ρ το? πατρ?ς, Plutarch Demetrius 51.2; Moralia 183C). This incident, however, is quite extraordinary, and it cannot be brought into a more general context. The other episode is not really of the same type as the previous examples; in 361/0 B.C. Dionysius the Younger persuaded the Pythagorean Archytus and his followers to secure Dionysius’ promise to safeguard Plato in Syracuse (Plutarch Dion 18.2; 20.1). As the vocabulary of this anecdote is different from the others,  it is perhaps best to do no more than narrate it.
Sometimes, however, a recipient did try to dictate the choice of hostages, usually because the donor sought a favor with particular desperation.  The conference which Antigonus requested with Eumenes  in 320 B.C. (Plutarch Eumenes 10. 2-3),  the deal which Agathocles tried to negotiate with Deinocrates in 306 B.C. (Diodorus 20.79.4), and the aid which Cleomenes of Sparta in 223/2 B.C. asked of Ptolemy III Euergetes of Egypt (Plutarch Cleomenes 22.3-7) and the Messenians of the Achaean League in 219 B.C. (Polybius 4.9.5) all indicated to the recipient that the donor could not refuse to grant him his choice since the donor could not escape from a very difficult situation without the recipient’s cooperation.
Although frequently in private agreements and regularly in treaties the recipient determined the choice of hostages, opportunities for deception remained. Plutarch and Macrobius recount that in 389B.C., in revenge for the rape of the Sabine women, the Latins demanded freeborn Roman women “for marriage alliances” from a Rome exhausted by the Gallic invasion (Plutarch Romulus 29.4; Camillus 33.2-7; Macrobius Saturnalia 1.11.35-40; cf. Plutarch Moralia 313A). Afraid to refuse, the Romans escaped humiliation first by the substitution of slave women and then by their successful surprise attack upon the enemy. The labored parallels and the general implausibility of the tale need no elaborate refutation; this is not an historical but a literary anecdote. But a comparison with the story of Popaedius Silo’s substitution of slave children and lead-plated goods in place of his own legitimate children and solid plate (Appian Civil Wars 1.85) produces an interesting question: could a recipient be deceived? There could be no point in deception if the donor sincerely meant to uphold his part of the conditions, for revelation of such a deception could probably negate the agreement. In both cases of substitution it is significant that the deception was meant as a temporizing measure until some resistance could be prepared,  while the recipient was lulled to a false sense of security by the donor’s apparent compliance. Was it possible to avoid giving the stipulated hostages by such means? Theoretically, it might seem, under certain circumstances, the obvious thing to try. Practically, the recipient’s knowledge of the donor’s circumstances, the strong possibility of betrayal by an informant - whether domestic, a personal enemy, or a foreign agent  - and the difficulty of an adequate impersonation over a prolonged period militate against the success of deception.
Thus the selection of hostages was a variable factor which depended upon the relative strengths of donor and recipient. Yet it is almost certain that the ancient evidence, which probably reflects only the final stage of bilateral agreements, has omitted preliminary proceedings: the proposal of certain criteria, the acceptance or refusal of those criteria, the production of a list of names which was then either approved or rejected. One source, unfortunately the late and not altogether reliable Plutarch, has preserved an anecdote of the supposed negotiations between Antipater and Sparta in 331 B.C.; Antipater’s demand for fifty boys was met with a counterproposal of one hundred old men or women (Plutarch Moralia 235 B-C). Although the historicity of the incident is dubious,  we may infer that there existed the possibility of selection according to mutually agreeable criteria.
References to time in formal agreements fall into two categories: the period granted for the delivery of the hostages and the  term of service for the individual hostages and for the entire contract. The former category needs no explanation, since it is obvious that the donor state would have required some time to notify and transport the chosen persons, especially if the donor state covered a large geographical area (Caesar Gallic War 4.27.6, 7.12.3-4). Moreover, a delivery deadline is equally advantageous for the recipient state because it establishes a definite date and act by which the recipient may judge whether the donor state has accepted the terms of the agreement. Although relatively few truces and treaties are specifically said to include such a clause (Cato’s Spanish settlements in 195 B.C. [Polyaenus 8.17.1]; the surrender of Carthage in 149 B.C. [Polybius 36.4.6; Appian Punic Wars 77]; Caesar’s demand to the Pirustae in 54 B.C. [Caesar Civil War 5.1.8]; and Vercingetorix’s revolt in 52 B.C. [Caesar Gallic War 7.64.11]), it must have been a standard feature in all negotiations. The period of delivery is very imperfectly attested as “at a certain day” (Polyaenus 8.17.1; Caesar Civil War 5.1.8; 7.64.1), within thirty days (Polybius 36.4.6), or immediately (Caesar Gallic War 4.27.6). 
The evidence for the delivery date of hostages is scanty but straightforward. The second matter (i.e., the term of service), however, is full of problems, and the ancient authors seem to mention at least three possibilities which the single topic “length of exaction” covers: 1) detention of hostages as guarantees for peaceful conduct for a fixed period of time; 2) detention of hostages as guarantees for peaceful conduct until the performance of a certain act; 3) the term of service for individual hostages.  To the first category belongs the Aetolian Treaty of 189 B.C., which limited the hostages  to six years’ service (Polybius 21.32.10), a term identical to the period over which the war indemnity was to be paid. The remaining evidence may be organized around two very different questions.
The first devolves upon the fate of those hostages who guaranteed an armistice which did not finally culminate in a peace settlement. Were such hostages returned to their own government or were they detained by the recipient state? In the case of four armistices our sources incorporate information on this matter. The clearest is that between Flamininus and Philip V of Macedon in 197 B.C., when Flamininus granted Philip a four months’ truce for the purpose of negotiating with the Senate a peace treaty; in return, Philip delivered hostages (including his younger son Demetrius) and two hundred talents, which were to be returned if Rome failed to ratify an agreement (Livy 33.13.14-15; Polybius 18.39.5-6; Appian Macedonian Affairs 4.2; Zonaras 9.16). In a similar incident, the conditional clause of the agreement has not survived, but the events demonstrate that it must have existed; in 151 B.C., the Roman general Marcellus repatriated the hostages of the Celtiberians after the Senate refused to ratify the preliminary terms (Appian Wars in Spain 50). The other two armistices require scholarly interpretation, and for that reason they have been used variously, both to support and to contradict the two preceding examples. In 83 B.C., during the civil war between the Marian generals and Sulla, recently returned from his victories in the East, the Marian L. Scipio received hostages from Sulla as pledges for a truce during which they were to discuss a settlement of their  disputes; when Scipio’s colleague Sertorius violated the truce by seizing Suessa, Scipio released the hostages without having been asked to do so (Appian Civil Wars 1.85).
If our information ended there, the implication would be clear: the termination of the truce released the hostages from their obligation. But the narrative continues and Appian relates that the capture of Suessa and the return of the hostages, which Sulla had not sought, alienated Scipio’s troops from him, and they then deserted en masse to Sulla. If Appian has not misinterpreted the episode by assuming the defection was motivated by the truce violation, there are two possible explanations for this disaffection among Scipio’s soldiers: either they felt disgusted that Scipio had forfeited through treachery or stupidity so obvious an advantage, or they believed that Scipio and the other Marians did not wish to cease hostilities under a negotiated peace. In the case of the former, the disaffection indicates that release of the hostages was not the expected result of the truce violation, and in the latter that the release demonstrated the indifference of the Marian party to continued negotiations; the latter makes better sense of the curious emphasis on the fact that Sulla had not requested their return. 
The passage which has generated by far the most argument is a paragraph describing Scipio’s preliminary conditions for a settlement after the Battle of Zama. Appian (Punic Wars 54) had put into Scipio’s mouth a speech in which 150 hostages of Scipio’s choice, 1000 talents, and provisions for the Roman army would purchase an armistice while an embassy to Rome was prepared; should the Senate ratify the treaty, the hostages were to be released (κα? γενομ?νων τ?ν σπονδ?ν ?πολ?ψεσθε  τ? ?μηρα). Several questions have been posed by those who have compared this passage with the other accounts of the preliminary treaty (Polybius 15.18.8; Livy 30.37.6), but only two of these questions need concern us here. First, at what stage in the negotiations were the hostages delivered? Second, what is the meaning of the ratification proviso?
The former query can be more cogently expressed in terms of what the hostages guaranteed. As the passage implies, the hostages must surely have guaranteed the peaceful conduct of the Carthaginians while their legation negotiated final terms in Rome,  and so it is reasonable to assume that these hostages were delivered almost immediately to Scipio. A comparison of this account with that of the Polybian-Livian tradition, however, discloses three points of disagreement; the total number exacted, their ages, and their purpose.  For Polybius, these hostages guaranteed the final peace. Since Polybius essentially identified the preliminary with the final ratified terms, and since Carthaginian hostages remained in Roman hands until at least 168 B.C. (Livy 45.14.5), Polybius’ interpretation is certainly valid. Moreover, we have another instance in which Polybius seems to have conflated preliminary and final terms; although Demetrius of Macedon served as a hostage in Rome from 197 to 191 B.C., Polybius did not include this condition in his description of the final treaty with Philip (18.44) but mentions it only in the preliminary negotiations, when Demetrius guaranteed the armistice (15.39.5-6). Similarly, Livy’s presentation of the campaign against Nabis (34.35.11; 40.4), which is based on Polybius’ narrative,  also describes the  exaction of the hostages from Sparta as pledges for the truce between Nabis and the Romans, and we may infer that he then simply omitted repeating these terms in the final treaty. Thus it should be clear that the discrepancies between the accounts in Appian and Polybius concerning the Treaty of Zama and its preliminary armistice in no way discredit Appian’s testimony, nor are the two accounts irreconcilable.
The ratification proviso is more problematic. Deceptively reasonable at the first reading, it states that the hostages arc to be released on the Senate’s ratification of a treaty, which seems to be consistent with the assumption that they guaranteed the truce only; after ratification, the truce would no longer have pertained. But Appian did not refer to any subsequent exaction, and Carthaginian hostages were still furnished to Rome a third of a century later. Only two explanations appear to me to be possible: 1) Appian is correct but for some reason neglected to mention the second exaction; or 2) Appian is wrong. The first premise leads to an additional problem, the inefficient and clumsy second exaction. If the persons involved in the two exactions were substantially the same, their release and then second delivery to the Romans is incomprehensible and pointless. But if the two groups were substantially different, it is difficult to understand both why the first group was unsuited to further service after only a few months’ time, and from where Carthage would have obtained so large a pool of youths of the highest nobility, after the recent depletion of Carthage’s population of military age. 
Moscovich rightly perceived the difficulty of the received text when he asked what would have happened had negotiations failed.   If negotiations during a truce lead to a final peace treaty, there is no need for the armistice to have settled the fate of the hostages for the period of time covered by the final peace; their release or further detention after the time of the truce may be included among the provisions of the final treaty. There is every need to determine their fate if negotiations fail, because there is no later opportunity for a discussion of the matter. Moreover, the purpose of a truce is to preserve the status quo. Possession of hostages shielded the recipient state from ambushes or surprise attacks by the donor;  a return clause would have protected the donor state from threats against the hostages after the truce expired. If the hostages indeed guaranteed Carthaginian good behavior during the truce, and the truce had ended in a failure to establish peace but without any violation of the armistice’s conditions, could Scipio have retained the Punic hostages? Retention was certainly expedient, as Aymard observed. His contention that the restoration of the Carthaginian hostages would have been folly for the Romans because it would have reestablished the armistice situation, to the benefit of the Carthaginians,  disregards both the legality of the case and the delicate position of Scipio and the Senate.  Nor is it obvious how reestablishment of the pre-armistice situation otherwise benefited the Carthaginians significantly; Scipio’s army still held major portions of their territory, and even if the hostages were returned, the provisions and the money which paid the Roman troops could not be given back. Aymard postulated that the desperate plight of Carthage after the defeat at Zama obviated the necessity for the return clause natural in a negotiated  armistice and that such a concession would have been “misplaced generosity” in the victorious general’s terms to a defeated people.  This argument greatly exaggerates the strength of the Roman military position and ignores the complexities of Roman domestic intrigue. Scipio’s fear of replacement, the exhaustion of Rome and Italy, the resistance of which Carthage, a strongly fortified and armed city, was still capable,  all militated against any prolongation of the war in order to insist on unconditional surrender.
Täubler decided that Appian (or his scribe) erred and emended Appian’s text by the insertion of a negative adverb: κα? ο? γενομ?νον τ?ν σπονδ?ν ?πολ?ψεσθε τ? ?μηρα.  Walbank’s objection to ο? is commendable;  μ? is preferable, but this is not a substantial criticism. Aside from the reservations which always attend an emendation of the received text when it is even slightly intelligible, there seem few explicit arguments against it. Aymard maintained, as we have seen above, the impracticality of Scipio’s release of Punic hostages for the truce but the inefficiency of, and lack of evidence for, a second exaction (both demonstrated by Moscovich) indicate the weakness of Aymard’s position. Täubler’s addition of a negative adverb has, moreover, the advantage of parallel incidents.  Nonetheless, the emendation has not met with general acceptance, despite what I feel is the best solution to this difficult and perhaps insolvable problem.
Appian’s account of the 300 Carthaginians at the outset of the third Punic War (Punic Wars 77) also supports the hypothesis that some conditions concerning repatriation were formally included in armistices and some treaties.  The lack of such conditions was considered suspicious  by the Carthaginians and contributed to their uneasiness about Roman demands. It is of course possible that Appian contributed this detail from his rhetorical training rather than from his historical research. But no release of these hostages ever occurred; they lived out their lives in honorable confinement in Italy with other Carthaginian prisoners (Zonaras 9.30). 
The second question which arises from the surviving material on temporal regulations is a serious and important problem which we cannot answer: for how long did a donor state provide hostages? As noted above, the Aetolian Treaty of 189 B.C. specifies six years as the period for which hostages were submitted and the indemnity paid (Polybius 21.32.10). Did all donor states continue to furnish hostages only until the full payment of the indemnity?  We know that the Treaty of Zama established that Carthage pay fifty annual installments, and Punic hostages remained in Roman control until at least 168 B.C. (Livy 45.14.5). Philip V of Macedon received the release of his son Demetrius and the cancellation of his unpaid debt to the Roman treasury as a double gift in return for his aid to the Roman army (Polybius 21.3.3; Livy 36.35.13; Eutropius 4.3.3.; Appian Syrian Wars 20; Zonaras 9.19). Demetrius of Syria claimed that all legal pretext for his detention had disappeared by 164 B.C. (Polybius 31.2);  he had come to Rome in 175 B.C., and the last indemnity installment was paid in 173 B.C. (Livy 42.6.6-7). Can we assume that the obligation to supply hostages terminated with the final indemnity payment? I believe that the hypothesis is not unreasonable; but the evidence is too sparse to make such a hypothesis more than speculative. 
Thus we may conclude that temporal regulations undoubtedly occurred far more often than the extant material indicates. An initial delivery date and some sort of expiration clause must have been included in almost every formal agreement, but the very frequency of such clauses probably made them seem routine and so contributed to the failure of ancient writers to relate them.
The formal provisions for the exchange of hostages (mutatio obsidum) involve considerable uncertainty. Only one treaty actually describes the procedure, that of the Romans with Antiochus III the Great of Syria in 188 B.C. (Polybius 21.42.22; Livy 38.38.9; Appian Syrian Wars 39). All three sources agree that the exchange was triennial, and Appian supplies the additional details that Antiochus, the son of the king, was excluded from the exchange and that the Senate, not the consul, had added the entire stipulation at the time of the treaty’s ratification. There seems no reason to dispute either point. Unfortunately the only exchange of hostages pertaining to this treaty is irrelevant. When Demetrius, the son of Seleucus IV Philopator, replaced his uncle Antiochus at Rome in 175 B.C. (Appian Syrian Wars 45), Seleucus was attesting his personal guarantee for the treaty’s continuance. Since, by the terms of the treaty of 188 B.C., Antiochus was the only hostage exempt from replacement, it is impossible to consider Demetrius’ substitution for him in 175 B.C. as part of the formal conditions of the treaty of 188 B.C. The death of Antiochus III the Great in 187 B.C. had dissolved the agreement, and Seleucus renewed the compact.  Whether the exchange clause was also renewed can only  be guessed at.
The Roman-Aetolian treaty of 189 B.C., however, contains two stipulations deserving some attention in this context (Polybius 21.32.11). The first is the specified duration of the service of the Aetolian hostages, six years, which coincides with the number of yearly installments to be paid on the war indemnity.  The second is the replacement of any hostages who should meanwhile die. In view of the youth of the selected hostages (twelve to forty at their selection, and thus eighteen to forty-six on their restoration, Polybius 21.32.10; Livy 38.38.6) and the short period of detention, this replacement clause is a very cautious, conservative measure. It appears reasonable to assume that similar clauses were inserted in other treaties, especially when the detention period was greater.  Moscovich has rightly and convincingly argued that no reference to mutatio obsidium can be understood from this treaty.  Still, the obvious utility of a provision for the replacement of hostages who die, despite its peculiarity to the Aetolian treaty, suggests the possibility that the ancient sources may have omitted clauses of little importance either because they were standard and customary (and therefore uninteresting) or because the circumstances under which they were to become operative did not arise. Whatever the case, the lack of evidence in other treaties does not mean that some arrangement for exchange, either by formal stipulations or by custom, should not be understood as a standard feature of such treaties.
Again, the most significant episode for this type of clause is the Treaty of Zama in 201 B.C. The annalistic sources of Livy  describe two occasions, one in 199 B.C. and the other in 181 B.C., on which hostages were restored to the Carthaginians (Livy 32.2-4; 40.34.14).  The reliability of these sources has been seriously impugned, and the occasions have been entirely discounted.  It is true that some of the annalists preferred creative ficti on to historical fact, but it is equally true that others were fairly accurate historians, and this is precisely the sort of technical information which is likely to have been preserved in the official records. The mere absence of supportive evidence from independent sources, Polybius in particular, ought not be lead us to dismiss the possible merits of annalistic material without a hearing.
In each passage it is said that one hundred hostages were restored. The verb used is reddere, in which there is no unequivocal denotation of “exchange” in the sense needed here.  Immediately one recognizes the difficulty of restoring 200 persons when only 100 or 150 hostages were exacted by the treaty of 201 B.C.  Moscovich, however, has observed that the hostages delivered to the Romans for the violated armistice of 203 B.C. (Polybius 15.8.7) could be grouped with the other Carthaginians held in Italy; if so, reddere could mean outright restitution.  Thus we are not obliged to recognize as evidence for mutationes the “restorations” of 199 and 181 B.C., but a subtler factor may convince us that mutationes took place. Carthaginian hostages were detained in Italy until at least 168 B.C., when Masgaba, Masinissa’s son, demanded that Hanno, son of Hamilcar, be substituted for some other hostage (Livy 45.14.5). At that time hostages exacted according to the age requirements in 201 B.C. would  have been forty-seven to sixty-three years old and could hardly be expected to survive many more years. It seems unlikely that the government which carefully stipulated replacements for the Aetolian hostages who died during the six years of their detention would permit the gradual loss of Carthaginian hostages through death and increasingly poor health over thirty-three years. Moreover, the value of a hostage who was absent from Carthage for a third of a century would almost certainly have been much less than that of a hostage who had been in the city more recently; the personal bonds between the hostages and their friends and relatives would significantly weaken, thereby seriously impairing their value as hostages. The inevitable shifts in political influence of the oligarchic families posed a further danger, if hostages were taken only from families important in 201 B.C., for their influence may have been eclipsed by that of others not related to the Carthaginians in Roman control.  We may also infer that Masgaba’s proposal to have Hanno sent to replace some other hostage was apparently a reflection of Masinissa’s attempt to restrain the anti-Numidian policy of Hamilcar, Hanno’s father.
Clearly some provisions for exchange were necessary, and it is possible that the “restorations” of 199 and 181 B.C. were in fact exchanges.  How often and at what intervals might exchanges have taken place? The years 201, 199, 181 and 168 do not permit us to infer the existence of any regular system except that of annual replacement, which is improbable because of the large number of persons involved if no duplication was allowed.  It would have been impossible, if there were one hundred new hostages every year, to maintain a principle  of selection from the best families, since the age limit was restricted to men aged fourteen to thirty years old, and Carthage was governed by relatively few families. Moreover, the administration of details such as selection, transport and quartering for frequent exchanges would have been extremely cumbersome. Aymard is probably correct in asserting that replacement was irregular rather than “a solidly established usage” codified in custom if not in the treaty itself. 
Two other passages appear to concern the practice of mutatio obsidum. After Opimius’ victories on behalf of Massilia in 155 B.C., he compelled the Ligurians to supply the Massiliots with hostages κατ? τινας τακτο?ς χρ?νους (Polybius 33.10.12). The tantalizing plural χρ?νους tempts one to consider it probable that the treaty contained a formal exchange procedure, but the lack of further explicit evidence prevents one from having more than suspicion. Phillipson has classified this incident with examples specifying the time allotted or the delivery of hostages,  but the plural puzzles me.
Describing Augustus’ moderation toward barbarian chiefs, Suetonius wrote: et tamen potestatem semper omnibus fecit quotiens vellent obsides recipiendi (Augustus 21.2). Aymard has admirably discussed the difficulty in translating recipiendi as “taking back”; if the barbarian leaders could seek restoration of their hostages, surely they would do so once for all as soon as possible.  There would be no point in the exaction if hostages could be immediately reclaimed. In this context the verb recipere is the active equivalent of the passive reddi, and the confusion which Suetonius’ remark has produced is due as much to its brevity as to his assumption of his  readers’ knowledge of the general practices concerning hostages.
Hostages From Non-contracting Parties
It is perhaps ironic that a provision concerning hostages which embodies one of the most logical peace terms has left almost no surviving evidence for its existence. When one nation has ceded to another its claim on some piece of land, it is well-nigh inconceivable that the first nation should be allowed to keep hostages from the ceded area, for in that case, the loyalty of the inhabitants to the new ruler would undergo a severe strain from the outset. Caesar clearly understood that the Aeduan hostages in the power of Ariovistus and the Germans constituted a distinct threat to Roman authority in southern Gaul, and his presentation of this pretext for his German campaign of 58 B.C., must have been amply appreciated by his contemporaries (Gallic War 1.33.2).  Other examples of the repatriation of hostages are not uncommon, but such restorations were usually due to the fact that the hostages had been captured by a third party (i.e., neither the donor nor the recipient).  Only three passages, to my knowledge, specifically treat the release of hostages from a non-contracting state as part of the formal negotiations for peace, and they refer to separate events in different authors. Two are incontrovertible. In 201 B.C., the Carthaginians agreed to release the hostages whom they had been keeping from states that were now, as a result of the new treaty, outside the new boundaries of the Carthaginian empire (Appian Punic Wars 54).  Second, Pharnaces was to return to Ariarathes all the places which had formerly been in Ariarathes’ territory and the hostages therefrom, in accordance with the treaty of 180/179 B.C. (Polybius 25.2.6).  
The third passage has generated much more discussion because of its singularly problematic character. Nonius Marcellus in his compendium of unusual Latin forms preserved some lines from Naevius’ Punic War. id quoque paciscunt, moenia ut sint quae Lutatium
reconcilient; captivos plurimos idem
Sicilienses paciscit obsides ut reddant
 The last line is the crux of this discussion; the textual variants and emendations for the first two lines need not concern us here.  The role of idem, however, is important for an understanding of the line. Is it a part of the Naevian text, functioning as the nominative pronoun which is the subject of paciscit, or is it Nonius’ reference to Naevius, introducing a second citation from the same author?
If Warmington’s punctuation of the passage in Nonius is correct, it must follow that idem refers to Lutatius (or, less plausibly, to another Roman negotiator) and that the lines belong to the peace terms of 241 B.C. which, we must assume, were described in Book VII. Warmington postulated the Carthaginians as the subject of paciscunt and the Sicilians as the subject of reddant.  Thus the Roman general would have stipulated that the Sicilians repatriate the Carthaginian prisoners of war who were held as hostages.  Several objections have been raised to this hypothesis, most notably by Täubler.  With the exception of Hiero’s domain, the Romans controlled Sicily, and the repatriation of Carthaginians held prisoner in Sicilian cities  could not have been the subject of an agreement with Hamilcar. Normally, the prisoners of war of a defeated nation were ransomed, detained as prisoners, or sold as slaves.  Moreover, since the armies of Carthage were primarily composed of mercenaries, it is difficult to comprehend why Lutatius demanded that the already financially embarrassed city significantly add to her troubles by releasing mercenaries who would return to demand wages. The hypothesis of Leo and Warmington is seriously undermined by the complexity of a single line’s reference to three nationalities, by the lack of evidence for Carthaginians under Sicilian control at the end of the First Punic War, and by the fact that such an agreement would have contradicted the usual practice in the matter of the prisoners and/or hostages.
If Nonius inserted idem to signify the beginning of another Naevian quotation in which the active verb pacisco was used, it is no longer necessary to supply Lutatius as the subject of paciscit. Cichorius on this basis proposed a quite different interpretation of the fragment, citing as further evidence for the separation of the passage into two fragments the alteration of the plural paciscunt to the singular paciscit.  He argued that Sicilienses modified the accusative obsides, which suggested to him as the understood subject of paciscit Hiero II of Syracuse. He maintained that only Hiero could have given Sicilian hostages and that Hiero was the sole Sicilian candidate for the singular verb in this context. The hostages would have been exacted by the treaty made with the Romans in 263 B.C. as guarantees for the fifteen installments of the war indemnity levied on  Syracuse.  This line, Cichorius believed, belonged to Hiero’ s request for their return by the Romans after the last installment was paid in 248 B.C. and would therefore be more properly assigned to Book VI than to Book VII.
While Cichorius’ hypothesis is supported by a suitable occasion for the exaction of Sicilian hostages, no other evidence corroborates its assumption that hostages were in fact given to the Romans by the Syracusans. Nor has Cichorius adequately explained why a fragment of Book VI should follow, rather than precede, a fragment of Book VII; Nonius tended to preserve the order of his citations,  which inclines one to suspect that this line not only belonged to Book VII but also followed the other line(s), although possibly after an interval.  The verb paciscit also seems peculiar. The agreement was contracted in 263 B.C., and unless the verb refers to a renewal of the Roman-Syracusan alliance on condition that Hiero’s Sicilian hostages were returned, it is difficult to understand it in its usual meaning.  In view of Hiero’s care to sustain a cordial, perhaps even a fawning, subordination to his Roman allies, a request for the repatriation of hostages before the end of hostilities is uncharacteristic; the Romans might well suspect treachery if Hiero seemed too eager for their release. 
The interpretation offered by Täubler best settles the problems raised by this perplexing passage.  The subject of paciscit is Lutatius, that of reddant the Carthaginians, and Sicilienses modifies obsides. Thus the Roman general sought concrete demonstration of the Carthaginian concession of Sicily by stipulating that the Sicilian hostages be restored. How had the Carthaginians obtained Sicilian  hostages? Täubler proposed that the Carthaginian generals campaigning in Sicily would have taken hostages from whatever territory lay in their control to prevent defection to the Romans and to safeguard themselves from guerrilla attacks. Again, however, there is no positive evidence to prove that the Carthaginians, who had taken hostages from Hecatompylus (Theveste) in Numidia in 243/2 B.C. (Diodorus 24.10.2) and from Syracuse in 289 B.C. (Diodorus 21.18), exacted Sicilian hostages during the First Punic War,  but that they did so seems a reasonable assumption.
The many difficulties inherent in trying to glean as much information as possible from so fragmentary a work as Naevius’ Punic War preclude any final solution of all the questions surrounding this line. Nevertheless, a fairly strong case has been made by Täubler which would group this fragment with the restorations of hostages in 201 and in 180/179 B.C. (Appian Punic Wars 54; Polybius 25.2.6);  the Naevian incident is thus not without parallel.  The logical necessity of demanding the release of hostages from surrendered territory very early in the negotiations may plausibly account for the scarcity of references to such a practice; the condition was probably fulfilled immediately, before further terms were contracted, and so it was rarely recorded in the official final agreement.
What, finally, may be postulated about the “typical” regulations concerning hostages in Roman practice?  On the basis of the evidence presented in this chapter it is possible to reconstruct the general characteristics of most of the hostages exacted by Rome in  the Republican period and to offer a list of clauses which in all likelihood were routinely employed whenever suitable circumstances arose.
Hostages taken according to a ratified treaty were usually young men, in the sense of the Latin word adulescens;  selected by the Roman magistrate in whose provincia the territory of the co-contractant lay, they were chosen from the highest-ranking families of the governing class. Their number depended on the quality of the hostages provided, and quality was a function of kinship and/or personal tie to the actual ruler or member of the ruling class. It is reasonable to assume that treaties typically included clauses in which the length of the hostageship of individuals and of the exaction as a whole was specified, and that provisions governing substitutions and restorations were clearly spelled out, as well as the procedures for any other foreseeable contingencies.
If we compare several Roman treaties, we can reconstruct a composite agreement which may allow us to consider hostage regulations in relation to each other and as a unified group. Such a composite agreement might include the following conditions: the government will give to the Romans x males as hostages; the hostages will be between twelve and forty years old; they will serve as hostages for y years which will either be the complete period for which the donor is to provide hostages or the period for which the individual hostages serve; they will be chosen by the Romans; exceptions may be made for certain magistrates and former hostages to Rome;  hostages will be provided for a total of z years/until a certain event occurs; then they will be released; if a hostage dies, another is to replace him; they will be delivered by a certain date and to a certain place or a certain official.  Of course, as seen in the discussion of the individual clauses, there was a great deal of flexibility within each limitation. Nevertheless, despite the failure of every extant treaty to include all eight conditions, I believe that these eight conditions must have been determined by negotiations between the contracting parties; although they may not have needed formal recognition in the treaty itself, some sort of agreement on these points would have been mandatory in almost every case.
Armistices involved somewhat different criteria. Since an armistice usually covered a very short period of time, the hostage regulations were not as stringent or as elaborate as those for treaties. The salient characteristics are: the government will give to the Romans x males as hostages; the hostages will be detained for the period of the truce, and for a longer period, if the peace settlement should require it; they will be delivered by a certain date and to a certain place or a certain official; if the donor government holds hostages from land which is to be ceded, those hostages will be released. Although other restrictions may also have been applied to the hostages who guaranteed a particular truce, such restrictions were less uniform than those which usually attended treaties; the short duration of service and the sometimes limited number of people available prevented  the employment of the full rigor of a formal exaction.
In both types of agreement, however, it should be clear that there were basic criteria which could be altered according to the particular situation. The Romans by no means adhered to an unchanging list of specific demands, but negotiated within a range of requirements.